10:00 - 10:30
10:30 - 12:10
10:30 -
Sentinel-1 Mission Status
Miranda, Nuno - ESA- ESRIN, Italy
10:50 -
COSMO-SkyMed Second Generation - Mission update
Dini, Luigi
Dini, Luigi - Italian Space Agency - ASI, Italy
11:10 -
TanDEM-X and Tandem-L: Mission Status
Hajnsek, Irena - DLR/ETH, Germany
11:30 -
The BIOMASS Mission
Scipal, Klaus - European Space Agency, Netherlands, The
11:50 -
On the Polarimetric potential of the STEREOID mission
Lopez-Dekker, Paco; Iannini, Lorenzo; Li, Yuanhao - Delft University of Technology, Netherlands, The
STEREOID (Stereo Thermo-Optically Enhanced Radar for Earth, Ocean, Ice, and land Dynamics) is one of the three mission proposals selected as Earth Explorer 10 candidates. If implemented, STEREOID will dramatically augment the capabilities of the Sentinel-1 mission by flying two identical sub-500 kg- class spacecraft carrying a receive-only radar instrument as main payload that will flying in a re-configurable formation with Sentinel-1D, which will be used as illuminator. STEREOID is conceived as a multipurpose mission, that will exploit its geometric diversity to help precisely quantify small scale motion and deformation fields of the ocean surface, glaciers and ice sheets, and solid Earth, aiming at providing modellers with data required to better understand dynamic processes in these three domains. The radar measurements will be enhanced and supported by a medium resolution dual VNIR and TIR payload. The mission will be organized in several phases, associated to distinct formation configurations. In our preliminary concept, the mission will start and end with a cross-track interferometric configuration, with both STEREOIDs flying in close formation, aimed at quantifying volume changes of ice masses and unstable grounds. In between, the two STEREOIDs would fly an extended period of time in a so-called StereoSAR configuration, with one spacecraft about 300 km ahead and the second one the same distance behind Sentinel-1. This configuration will maximize the sensitivity to surface motion vectors. STEREOID blends elements from the StereoSAR, SESAME, and PICOSAR mission concepts. An intriguing aspect of the STEREOID acquisition geometry is its polarimetric behaviour. As shown in literature, the azimuth bistatic angle introduces what can be understood as a rotation of the polarization basis. For example, for a rough surface at 90 degree azimuth bistatic angle, for a V-polarized transmit signal, the receive-polarization that would maximize the received power would be H. At 45 degree azimuth bistatic angle, STEREOID-like bistatic acquisitions would be polarimetrically analogous to a monostatic compact-polarimetric system transmitting a 45 degree linear polarized signal, which would allow retrieving most polarimetric information under the assumption of some common symmetry properties. However, For STEREOID, the azimuth bistatic angle is still a design parameter, but it can be expected to be in the range of 35 to 40 degree, in near range, decreasing to 20 to 30 degree in far range. In the final submission we will assess the polarimetric performance of this in some extent degraded compact-pol configuration. Another point of interest is the suppression of some scattering mechanisms by the bistatic geometry (e.g. scattering off dihedrals or trihedrals). Since STEREOID data sets will always include regular Sentinel-1 data, the polarimetric observation space will be extended from the two monostatic polarizations offered by Sentinel-1, to four or six (depending on the formation configuration) combinations of polarization and bistatic angle. We aim at providing a first assessment of how this extended observation space could be exploited.
12:10 - 12:30
12:30 - 13:30
13:30 - 15:30
13:30 -
Comparison of Hybrid Polarimetric Parameters from RISAT-1 and ALOS-2 Compact Mode SAR Data
Subrahmanyeswara Rao, Yalamanchili; Kumar, Vineet - Indian Institute of Technology Bombay, India
Many satellites, such as ENVISAT, RADARSAT-2, TerraSAR-X/TanDEM-X, Sentinel-1, with Synthetic Aperture Radar (SAR) system were operated or operating with dual-pol and fully-polarimetric modes for earth observation. After the launch of the RISAT-1 by ISRO and ALOS-2 by JAXA, compact polarimetric mode data are available to the users. RISAT-1 was the first Earth observation satellite to provide compact data operationally during 2012–2017. Japanese ALOS-2 is acquiring data in experimental mode over the Earth since 2014. Recently, Argentina launched SAOCOM-1A satellite with hybrid polarimetric mode. Future planned satellite missions such as SAOCOM-1B, Radar Constellation Mission (RCM) by Canada and NASA ISRO SAR (NISAR) will provide compact polarimetric data. Most of the work related to compact data is confined to simulated data from fully-polarimetric data. In this paper, we have compared the hybrid polarimetric parameters of RISAT-1, ALOS-2 real compact, ALOS-2 simulated compact from full-polarimetric SAR data and RADARSAT-2 simulated compact data. All the data sets were acquired over Vijayawada test site, India with a gap of one month in the year 2015. RISAT-1 compact and RADARSAT-2 full-pol mode data were acquired on May 24, and May 27 2015 respectively, whereas the ALOS-2 full-pol data and real compact data were acquired on April 1 and 15, 2015 respectively. Both RISAT-1 and ALOS-2 real compact data were acquired in right circular transmit and linear receive mode. In synchronous with the RISAT-1 pass, both trihedral and dihedral corner reflectors were mounted for the calibration of the data. With ALOS-2 passes, CRs were not mounted due to the lack information of the passes. All the data sets were acquired at an incidence angle of 36 to 39 degrees. RISAT-1 SAR data were processed for hybrid polarimetric parameters (ellipticity, relative phase, Pauli Phase, Axial ratio, CPR and degree of polarization) of CRs and distributed targets (bare fields, various crops and vegetation areas). Similarly, ALOS-2 real compact data and simulated compact data (converted full-pol data to compact) are also processed for the same hybrid parameters. The parameters are compared for the differences. We found some differences between RISAT-1 and RADARSAT-2 hybrid parameters as the RADARSAT-2 full-pol data is well calibrated. RISAT-1 data were calibrated for sigma-0, but the phase calibration is not done. There was also small offset between RH and RV channel data and channel imbalance between RH and RV. Due to this, a difference of 6 to 20 degrees from ideal value is observed in relative phase between RH and RV channel using CRs response, whereas for RADARSAT-2, these differences are 2 to 6 degrees. The difference in ellipticity angle is about 7 to 12 degrees for RISAT-1, whereas it is about 2 to 5 degrees for RADARSAT-2. Using distributed scatters, we observed differences in hybrid pol parameters between ALOS-2 simulated and real compact data. It may be noted that ALOS-2 full-pol data are well calibrated unlike to real-compact data, which is supplied without calibration. The results of comparison of RISAT-1 and ALOS-2 real compact data will be presented in this paper.
13:50 -
Comparison of signal models for change detection with polarimetric SAR
Marino, Armando - The University of Stirling, United Kingdom
Please see attached file for a 3 page pdf. The summary is in the following: In this work, two frameworks for change detection are quantitatively compared using real data and Monte Carlo simulations. The two methodologies are based on the use of the of Lagrangian optimisations of two distinct operators: a power ratio and a power difference. The two are based on different signal models, a multiplicative and an additive model. To compare the signal models extensive Monte Carlo simulations are performed. These clearly reveal that the eigenvalues of the multiplicative model produces a detector with higher power (higher probability of detection) while the eigenvectors of the additive model allow a straightforward physical interpretation of the dominant scattering mechanisms that have been changed in the scene. The two detectors are therefore complementary and their selection should depend on the specific application. We then proceed comparing the model using real data. For this, we used a large variety of data: ALOS-2 quad-pol data over urban, coastal and forested areas, RADARSAT-2 quad-pol over agricultural and coastal areas, Sentinel-1 dual-pol over coastal areas. The latter also allow to appreciate the different result between dual-pol and quad-pol data.
14:10 -
Biomass Estimation by means of Interferometric Ground Suppression in SAR Data
Mariotti d'Alessandro, Mauro (1); Tebaldini, Stefano (1); Quegan, Shaun (2); Soja, Maciej (3); Ulander, Lars M. H. (4) - 1: Politecnico di Milano; 2: University of Sheffield; 3: University of Tasmania; 4: Chalmers University of Technology
The continuous monitoring of the above ground vegetation has been greatly simplified since the introduction of dedicated remote sensing techniques. The analysis of both agricultural fields and unmanaged forests takes advantage of the very fast yet detailed coverage of huge areas provided by remote sensing. The problem of estimating the Above Ground Biomass (AGB) has been tackled by exploiting different kind of data: mainly optical, LiDAR and radar. This work focuses on the estimation of AGB in tropical forests by means of SAR measurements. This kind of forests are particularly interesting because they host most of the AGB on the Earth surface, but their irregular composition makes them very difficult to model. Recently an important result drove the attention on SAR tomography for the estimation of tropical AGB. It has been found that the backscattered power at P-band associated with 30m above the ground exhibits a very high correlation with tree biomass; most importantly, this correlation level does not saturate for AGB greater than 350T/Ha. A possible reason for this high correlation is that around 30m a small but constant percentage of the total biomass is found. According to this analysis 30m would be a perfect proxy for the total AGB; in addition, the P-band signal would be a good tradeoff between sensitivity to tree structure and penetration capability. Another possible explanation for this fact is that the backscattered power coming from the surroundings of 30m above the ground level is scarcely influenced by the ground echo. The echo coming from the ground carries information about the tree above, mainly because of the wave extinction and the double bounce scattering mechanism that concentrates the whole tree extension into one point placed at the feet of the trunk. Nonetheless it is also determined by many factors that are not related to tree biomass like ground roughness, moisture, ground topography that must accurately modeled if accurate AGB estimates are desired. TomoSAR 30m power is likely to overcome this modeling difficulties by simply rejecting the ground power. Should this last explanation hold then the problem that must be solved for estimating tropical AGB with P-band SAR data is how to effectively reject the backscattered power coming from the ground level. In this work a new technique that requires only two interferometric SAR images is presented: the interferometric ground notching. This technique is based on the subtraction between two SLC images whose phases have been accurately tuned: the information associated with the ground shared by the two images gets canceled whereas the contribution coming from the tree above gets emphasized. It is here shown how the effectiveness of the proposed approach is related to the normal baseline and to the knowledge of the ground topography; examples using real datasets are presented too. Results come from the processing of the data gathered by DLR and ONERA in the framework of the AfriSAR campaign, carried out by ESA as a support activity for the forthcoming BIOMASS mission. Reference biomass maps come from LiDAR measurements. For every dataset here analyzed a stronger correlation with biomass is observed after ground notching. Also, strategies for compensating the effects of the geometry of acquisition are described.
14:30 -
Polarimetric Two-Scale Model (PTSM): extension to anisotropic slope distribution and comparison with the second-order Small-Slope Approximation (SSA2)
Iodice, Antonio; Di Martino, Gerardo; Riccio, Daniele - Università di Napoli Federico II, Italy
The Polarimetric Two-Scale Model (PTSM) was introduced a few years ago as an electromagnetic scattering model to be used within algorithms for soil moisture retrieval from polarimetric SAR data [1-2]. PTSM inherits the ability to account for depolarization effects from the original Two-Scale Model (TSM) [3-4], and, with respect to the latter, it has the advantage to provide closed-form expressions of the elements of the covariance (or coherency) matrix, that hold for moderate large-scale surface slopes [1-2]. The TSM, also called Composite Model (CM), has been extensively used to study scattering from the sea surface, so that it is natural to explore the use of PTSM for the same purpose. However, in its current formulation PTSM assumes that the surface slope distribution is isotropic, which is not realistic for the sea surface. In fact, the variance of sea surface slope along the upwind (or downwind) direction is higher than the one along the cross-wind direction. Therefore, slopes along range and azimuth directions turn out to have different variance and to be correlated, and their variances and correlation coefficient can be expressed in terms of upwind and crosswind variances and of the angle between wind and ground range directions. Accordingly, we here extend PTSM to account for surface slope anisotropy, and this is the first contribution of the present work. In addition, as a second contribution, we provide PTSM expressions also in the circular polarization basis, which may be useful for some SAR sensor polarimetric configurations. A limitation of TSM and PTSM is that they only account for depolarization due to surface tilting, whereas they ignore depolarization due to multiple scattering, so that it is expected that they underestimate the depolarization effect. A more accurate model, that also accounts for the multiple scattering effect, is the second-order Small-Slope Approximation (SSA2) [5]. However, this higher accuracy is paid by the fact that SSA2 requires a computationally demanding numerical evaluation of fourfold integrals. Comparisons of numerical evaluations of TSM and SSA2 [5] show that generally they are in good agreement, except for the cross-polarized normalized radar cross-section (NRCS), which, in the considered cases, is underestimated by TSM of two to four dB with respect to the SSA2 value. However, since these comparisons were made for specific surface parameters, incidence angles and frequencies, no general conclusion can be drawn. Fortunately, in [6] an analytical approximation of SSA2 (SSA2-A) was obtained (although for the cross-polarized NRCS only), and it was shown that it is in good agreement with exact SSA2 for moderate slopes and off-nadir incidence angles. This allows us to obtain, as a third contribution of this work, an analytical closed-form expression of the ratio of cross-polarized NRCSs obtained by SSA2-A and PTSM: this ratio turns out to be a function of only surface dielectric constant and incidence angle. The obtained expression shows that for any values of these parameters (except for very near grazing angles) the difference of results of the two methods is not larger than about two dB. We can than conclude that, for applications in which computational efficiency is important (for instance, surface parameter retrieval algorithms) and small errors on the cross-polarized NRCS are acceptable, PTSM is preferable. REFERENCES [1] A. Iodice, A. Natale, and D. Riccio, “Retrieval of Soil Surface Parameters via a Polarimetric Two-Scale Model”, IEEE Trans. Geosc. Remote Sens., vol. 49, no. 7, pp. 2531-2547, July 2011. [2] A. Iodice, A. Natale, and D. Riccio, “Polarimetric two-scale model for soil moisture retrieval via dual-pol HH-VV SAR data,” IEEE J. Sel. Topics Appl. Earth Observ. Remote Sens., vol. 6, no. 3, pp. 1163–1171, Jun. 2013. [3] J. W. Wright, “A New Model for Sea Clutter”, IEEE Trans. Antennas Propagat., vol. 16, pp. 217-223, 1968. [4] G. R. Valenzuela, “Scattering of Electromagnetic Waves from a Tilted Slightly Rough Surface”, Radio Sci., vol. 3, pp. 1057-1066, 1968. [5] A. G. Voronovich and V. U. Zavorotny, “Full-polarization modeling of monostatic and bistatic radar scattering from a rough sea surface,” IEEE Trans. Antennas Propagat., vol. 62, no. 3, pp. 1362–1371, March 2014. [6] C. A. Guérin and J. T. Johnson, “A simplified formulation for rough surface cross-polarized backscattering under the second-order small-slope approximation,” IEEE Trans. Geosc. Remote Sens., vol. 53, no. 11, pp. 6308–6314, Nov 2015.
14:50 -
Incidence angle dependencies in narrow-swath Quad-polarimetric SAR images
Cristea, Anca; Doulgeris, Anthony Paul - University of Tromsø, Norway
SAR images are known to be affected by the incidence angle (IA) effect, which manifests itself as a decay of the average intensities with incidence angle. Therefore, the same material will appear brighter when viewed at near range than at far range. The decay constitutes a nuisance for both human observers and automatic image analysis algorithms such as segmentation (also known as clustering). In general, it is assumed that the effect is not a major concern for narrow swath quad-polarimetric scenes, due to the fact that they cover small incidence angle ranges. We have conducted an analysis on Fine Quad-Pol and Wide Fine Quad-Pol RADARSAT-2 scenes with the purpose of assessing the correctness of this claim. Numerous studies conducted on wide-swath dual-polarization SAR images have established that the log-intensities register an approximately linear decay as a function of the incidence angle, often expressed in dB/degree. The decay depends on polarization and also on surface roughness, therefore differs from one physical surface to another. In our studies, we focus on open water, sea ice and oil slicks. Based on these observations, we have previously developed a segmentation algorithm that deals with the incidence angle effect automatically and without requiring training. The distinguishing feature of the algorithm is the inclusion of the IA dependencies of the log-intensities into the basic Gaussian mixture model as a linear modulation of the cluster means. The algorithm is able to successfully eliminate typical wide-swath image segmentation artefacts such as banding and the general over-splitting of clusters, both of which appear in the range dimension. Additionally, as the dependency is considered to be linear, a slope (“decay rate”) /intercept value pair can be extracted for each cluster and channel and used as a feature for interpretation and classification. As the slopes depend on roughness, they are lowest over rough ice and highest over smooth areas such as water and oil films. RADARSAT-2 Fine Quad-Pol scenes cover an incidence angle range of up to 2 degrees, and Wide Fine Quad-Pol scenes cover a range of approximately 3 degrees. An analysis of both types of scenes containing water, oil and sea ice revealed that a noticeable incidence-angle dependency is present in the co-polarized channels. The slope feature can be extracted from the co-polarized channels of the narrow-swath images, with values similar to those estimated from wide-swath images. Because the decay is more pronounced for open water, the segmentation of scenes containing open water without considering the decay may lead to the appearance of bands in the range dimension, even for small IA ranges. After applying our algorithm, the banding effect is eliminated and the open water background is placed into a single cluster. The decay may however still be negligible in both Fine and Wide Fine Quad-Pol images of sea ice, where the brightness difference between clusters is more pronounced than the shift of the average cluster value due to IA variations. Different rates may still be measured for different ice types with highly variable roughness, but we have not yet encountered this case. It is important to also consider the variation of the slopes with the level of multi-looking. As the number of looks is increased, the individual clusters become better defined and their characteristic slopes are estimated with a higher precision and become more significant compared to their variance. Thus, low slopes may only become evident when the scene is heavily multilooked (e.g. 25x25 looks). In addition, segmentation can be performed at different levels of detail, with the slopes of some clusters becoming significant only at high levels of detail. Cross-polarized channels were not considered for the analysis, as they are affected by noise to an extent that prevents the extraction of valuable information concerning the intensity-incidence angle relation. For the moment, this proves to be a considerable drawback for automated segmentation, as many structures of interest such as ships or icebergs are more easily detectable in the cross-polarized channels. We conclude that an accurate segmentation requires accounting for incidence angle variations, however small, in particular for open water areas. The effect of the incidence angle may also be non-negligible in narrow-swath dual-pol, quad-pol or compact-pol images acquired from SAR sensors other than RADARSAT-2. We are currently only using intensity (backscatter) information, but we aim to also explore possible consequences for polarimetric matrices and extracted polarimetric features. In addition, we foresee the possibility of performing supervised or unsupervised cross-angle classification on datasets of narrow-swath images using the same modelling approach.
15:30 - 16:00
15:40 - 17:00
15:40 -
Relating temporal decorrelation at P and L bands: results from tower-based and airborne campaigns supporting future synergies between repeat-passes spaceborne missions
VILLARD, Ludovic (1); HAMADI, Alia (1); BORDERIES, Pierre (2); KOLECK, Thierry (1); LE TOAN, Thuy (1) - 1: CESBIO, France; 2: ONERA, France
Temporal decorrelation is known as a critical parameter for the design of spaceborne missions, even for the longest wavelengths currently operable from space or in a close future with the upcoming Biomass mission at P-band. In this framework, tower-based campaigns have been funded by ESA and CNES in order to obtain valuable time series from which temporal decorrelation can be calculated, considering time interval from 15 min to several months. Among key results supported by both TropiScat (2011-2013) and AfriScat (2015-2017) campaigns in French Guiana and Ghana respectively, the statistical analysis support the choice made for Biomass of helio-synchronous orbits at 6h and repeat passes shorter than 3 days. In addition to the P-band results, both campaigns also enabled to acquire time series at L-band with the same acquisition scenarios, hence the possible comparisons between temporal decorrelation for the two frequencies. Beyond the expected stronger temporal decorrelation at L-band, the gap with P-band results is shown to be closely related to seasonal effects, and non-linearly dependent on the temporal baseline. Supported by the physical basis of the MIPERS-4D forward model which accounts for dielectric changes and convective motion, a semi-empirical relation between temporal decorrelation at P and L bands has been derived as function of temporal baseline. Cross-validation between both TropiScat and AfriScat sites shows 95% confidence interval lower than 0.1, which strongly support the prediction of P-band temporal decorrelation from current or upcoming zero baseline L-band missions, and paves the way for future synergies across spaceborne missions.
16:00 -
Comparison of Nonlinear PCA and Model-Based Decompositions for Crop Classification using Multitemporal Fully Polarized SAR Data and Recurrent Neural Networks
Wagener, Nicolas Jakob (1); Vrinceanu, Cristina Andra (1); Del Frate, Fabio (2) - 1: European Space Agency; 2: University of Rome Tor Vergata
Crop type classification is an important application of remote sensing data, for example in the context of food security or agricultural policies. Neural networks have a potential for this purpose due to their ability of addressing nonlinear dependencies between measured backscatter and crop and soil characteristics. Recent studies have furthermore demonstrated the capabilities of long short-term memory (LSTM) recurrent neural networks (RNN) to properly handle dual-pol SAR crop time series and outperform more traditional machine learning approaches. However, an area that still needs to be further investigated is the extraction of pre-classification relevant features. In this study we use a fully polarimetric SAR time series for crop type mapping with an RNN-based classifier. The multilooked coherency matrices are decomposed using traditional model based approaches as well as an automatic decomposition approach based on nonlinear Principal Component Analysis (PCA). By comparing the different decomposition methods, this study shows innovative results on feature extraction for machine learning with polarimetric SAR data. Furthermore, it provides – to our knowledge for the first time - an investigation of fully polarimetric time series classification of crops using an RNN.
16:20 -
Confusion between polarimetric signatures of forests and slanted buildings at C- and L-bands. An explanation.
Thirion-Lefevre, Laetitia (1); Guinvarc'h, Régis (1); Colin-Koeniguer, Elise (2) - 1: SONDRA/CentraleSupélec, France; 2: DEMR/ONERA, France
The misclassification of urban areas at C- and L-bands is a classical problem in radar remote sensing. It is well established that when these areas are observed with a radar whose trajectory is not aligned with the main orientation of their streets, their polarimetric response in the Pauli basis signs as vegetation. Several studies have been dedicated to the correction of this misclassification. To our knowledge, none of them investigates the reason why this misclassification occurs. The purpose of this study is to explain the reasons for possible confusion between the polarimetric behavior of forests and urban areas. To do this, we first present the theoretical criteria, supported by electromagnetic modeling, which allows us to understand the main polarimetric contributions of the SAR signal over the city on the one hand and over the forest on the other hand. These considerations concern both the ratio of intensities between polarimetric acquisition chains, relative phases between co-polarised channels, as well as second-order statistical parameters such as entropy. We show that in some cases (metric resolution and oriented neighborhoods) none of these parameters make it possible to differentiate a priori between a vegetation zone and an urban zone. In a second step, we consolidate these general explanations by analyzing several polarimetric images, for various frequency bands, and types of resolution. Finally, in the last section we will propose alternatives for acquisition scenarios in order to remove potential ambiguities of misclassification. One of them relies on the use of the absolute value of HV.
16:40 -
Physical modelling of a multistatic coherence tensor in polarimetric SAR
ALVAREZ-PEREZ, JOSE LUIS - University of Alcala (UAH), Spain
Monostatic polarimetry in the theory of scattering of electromagnetic waves was initially reduced to the analysis of the vector Radiative Transfer equation of partially polarized radiation [TKS85]. Later on, several algebraic, matrix-based were developed [LP09]. On the other hand, and since interference phenomena are a specific phenomenon of waves, polarimetric interferometry is both governed by polarization and coherence effects. The former is related to similar issues as the ones relevant for monostatic polarimetry, where as new challenges arise in the definition and use of coherence. The use of sheer electromagnetic theory is sometimes complemented or even replaced by tools aimed at retrieving information from SAR polarimetric data with the use of paradigms that incorporate heuristics at some point [AP12, AP15]. Coherence is a descriptor of the electromagnetic fields as a vector field in the space domain containing the scattered fields and therefore must be analyzed in the sense described by Glauber. He introduced coherence as a mathematical separability condition for the correlation functions at all orders [Gla63], although the description led byWolf and Mandel [MW65] has had more use and consideration and has been the ingredient in polarimetric SAR interferometry. However, in the 70s Molyneux, McCoy and others (a review on the subject is provided in [Kra92]) investigated the mathematical equations arising from Maxwell and describing high-order correlation functions of the electric field at different points. In the work presented here, it is intended to continue motivating the interest in multistatic simultaneous data takes with SAR sensors at the light of the equations that extend the usual formulation for second-order coherence. To do that, we use the concept of coherence tensor as introduced in [AP15]. This extended concept of coherence makes it possible to incorporate multipoint measurements to form correlations of the electric field after scattering in composite random media, and it leads to the handling of tensors. It has been customary in SAR polarimetry to implement algebraic tools such as matrix eigenanalysis to facilitate the extraction of information about the scattering processes. In the extended version of coherence that captures the ideas of Glauber, the definition of a spectrum is more complex than for matrices, as explained in [AP15] and its introduction is justified by non-Gaussianity of the correlation among scattered fields recorded at multiple points. From another perspective, scattering can be analytically modelled with the hindrance of difficult inversion methods, without using polarimetric information to generate matrices with special symmetries. The combination of analytical modelling and matrix methods remains an open research subject. To relate both fields, and to incorporate smoothly the algebra of tensors, a dyadic Green’s function analysis is presented, in which two species of scatterers are considered, one representing volume scattering and the other rough surface scattering. The dyadic Green’s function in the k-space [AP06] is proven to be the most adequate candidate to model the scattering matrix in its target scattering vector representation, based on the Pauli basis. This Green’s function can be physically pictured with the help of Feynman diagrams as in [AP06] and this is the formalism chosen here to visualize the different contributions and orders of the scattering phenomena involved. It is in this formalism where the three tensorial spectral techniques described in [AP15] are now analyzed. In particular, the CANDECOMP/ PARAFAC decomposition’s rank is shown to describe the relative strength of the different scattering mechanisms better that a simple matrix analysis based on twopoint interferometric approaches. The Tucker decomposition reveals some internal structure and symmetries of the dyadic Green’s function describing the whole process of scattering in a composite random media of the type under study, namely rough surface and volume scattering together. Finally, the variational approach is the most adequate tool of the three to define a spectrum of the coherence tensor with some continuos and some discrete parts. The study is completed with the study of the connection of these methods with the spectrum of the dyadic as an operator in a Hilbert space and its decomposition into a point spectrum, a continuos spectrum and a residual spectrum. By showing the results of analytical modelling based on the small perturbation method together with volume-scattering renormalization methods as well as results from numerical simulations, we illustrate the interest and possibilities of multistatic simultaneous SAR measurements that allow the construction and analysis of the coherence tensor. In addition to this, some new light is shed on the well established techniques in polarimetric interferometry such as the random-volume-over-ground (RVOG), the interferometric water cloud model (IWCM) or the oriented-volume-over-ground (OVUG) model [Clo09]. [AP06] J. L. Alvarez-Perez. Renormalization of the Helmholtz equation for the problem of electromagnetic propagation in a layer of spherical scatterers. 17(2):103–119, April 2006. REFERENCES [AP12] J. L.Alvarez-Perez. The IEM2M rough-surface scattering model for complex-permittivity scattering media. 22(2):207–233, May 2012. [AP15] J. L. Alvarez-Perez. A Multidimensional Extension of the Concept of Coherence in Polarimetric SAR Interferometry. 53(3):1257–1270, March 2015. [Clo09] S. R. Cloude. Polarisation: Applications in Remote Sensing. OUP Oxford, Oxford, UK, 2009. [Gla63] R.J. Glauber. The quantum theory of optical coherence. Phys. Rev., 130(6):2529–39, 1963. [Kra92] Y. A. Kravtsov. Propagation of electromagnetic waves through a turbulent atmosphere. Rep. Prog. Phys, 55(1):39–112, 1992. [LP09] J.-S. Lee and E. Pottier. Polarimetric Radar Imaging: From Basics to Applications. CRC Press. Taylor & Francis Group., Boca Raton, FL, 2009. [MW65] L. Mandel and E. Wolf. Coherence properties of optical fields. Rev. Mod. Phys., 37(2):231–87, 1965. [TKS85] L. Tsang, J. A. Kong, and R. T. Shin. Theory of microwave remote sensing. Wiley Series in Remote Sensing. John Wiley and Sons, New York, 1985.
17:00 - 17:20
17:20 - 19:00
09:30 - 10:30
09:30 -
Application of polarimetric information to the atmospheric phase screen compensation for GB-SAR in a mountainous area
Izumi, Yuta (1); Zou, Lilong (2); Kikuta, Kazutaka (1); Sato, Motoyuki (1) - 1: Tohoku university, Japan; 2: The National Institute of Advanced Industrial Science and Technology (AIST), Japan
A ground-based polarimetric synthetic aperture radar (GB-PolSAR) system was installed for monitoring of landslide affected area in Kumamoto, Japan since 2017. Because the observed area is in a mountainous region, atmospheric phase screen (APS) artifacts are intense and linear approximation along the range is not effective due to the steep topography. In this case, the multiple-regression model is the proper method for compensating APS, but it needs enough coherent scatterers (CS) for accurate estimation. In this study, we investigate the potential of the full polarimetric information to the APS compensation for the operating GB-PolSAR. Specifically, owing to the fully polarimetric acquisition capability of our GB-PolSAR system, we apply polarimetric optimization and PolInSAR covariance matrix similarity test (similarity test) approaches to improve the accuracy of APS compensation. The APS is the most relevant artifact on the interferometric phase, which is classically compensated by model-based technique, applying to higher coherence pixels. In the APS estimation step, less number of CS degrades the compensated results because the APS is estimated by the squares regression. However, our GB-PolSAR monitoring area at a mountainous region is not always applicable to yield high coherence, especially in a heterogeneous region where sparse vegetation and bare surface soil are mixed. In this case, the estimation of APS becomes inaccurate. In order to obtain higher coherence along whole observed area including homogeneous and heterogeneous regions, we herein propose to apply polarimetric optimization combined with similarity test approaches in this study. Among polarimetric optimization algorithms, two methods which include; BEST method and equal scattering mechanism (ESM), were selected and evaluated. The similarity test evaluates the similarity of both polarimetric and interferometric property at the spatial averaging step of complex coherence. As a result of this averaging, higher coherence magnitude can be achieved at heterogeneous region and point target. We employ considerable aspect of both approaches to increase the complex coherence magnitude. We compared the APS compensated results between the single polarization (HH) and the full polarization, and also between the boxcar filtered and the similarity test applied datasets acquired by the GB-PolSAR system at the mountainous area. The polarimetric optimization yielded a higher number of CS, resulting in more accurate APS compensation. Furthermore, the increase of the number of CS was also confirmed at the heterogeneous area by applying the similarity test. Finally, the comparison results indicated that polarimetric optimization combined with similarity test revealed the most accurate compensated result.
09:50 -
PolInSAR Ground and Volume Response Separation
Alonso-Gonzalez, Alberto; Papathanassiou, Konstantinos P. - DLR, Germany
The radar response over vegetated areas is usually separated into the ground and volume contributions. Polarimetric SAR (PolSAR) has the ability to detect different scattering mechanisms within the same resolution cell. Therefore, several decomposition techniques have been proposed in the literature to separate the two responses from PolSAR data. However, this separation is an undetermined problem and several assumptions need to be performed. In this regard, Polarimetric SAR Interferometry (PolInSAR) has the additional advantage of having also sensitivity to the vertical structure and distribution of the scatterers. For this reason, some future SAR missions like BIOMASS or Tandem-L will have a strong focus on acquiring PolInSAR and tomographic time series over forested areas. Some PolInSAR models as the Random Volume over Ground (RVoG) have been used in the literature in order to estimate some vegetation properties as height or extinction [1]. This work describes a PolInSAR technique, based on the same two layer model and formulation, to separate the ground and volume responses. Accordingly, the fully polarimetric covariance matrices of the ground and volume components are extracted from single or multi-baseline data. The modelled vertical profile consists of a layer of volumetric scattering on top of an impenetrable ground. This assumption results in the coherence linearity, as already described in [1], where the observed coherence region follows a line segment between the ground and volume coherences, determined by the vertical structure of each layer. In [2] the mathematical constraints for the coherence linearity were derived. Accordingly, the coherence region should be the affine transform of a Hermitian matrix. Here, the same idea is extended in order to describe the full PolInSAR response in terms of the ground and volume coherences and covariance matrices. A multi-baseline approach has been proposed in [3] and it has been shown that over forest the estimated volume and ground covariance matrices fulfill a combination of a random volume and a surface plus dihedral. Moreover, most of the details over the forest areas corresponding to the underlying terrain appear in the ground component. In this work, single-baseline and multi-baseline approaches will be proposed and compared, in the direction of future missions like BIOMASS and Tandem-L. Special attention will be given to the model assumptions and limitations. A stability and baseline dependency analysis will be included, in order to give an idea of the technique performance. To perform this evaluation, real data at L-band and P-band over different types of forests acquired by the DLR E-SAR and F-SAR sensors will be employed. Ground truth information like ground and forest height extracted from LIDAR measurements is available over most of these datasets. This gives us a good opportunity to analyze the link between these physical parameters and the obtained ground and volume components. The extracted ground and volume components will be analyzed in detail, comparing their polarimetric information with traditional models employed to describe these scenarios. [1] Cloude, S. R., and Papathanassiou, K. P. (2003). Three-stage inversion process for polarimetric SAR interferometry. IEE Proceedings-Radar, Sonar and Navigation, 150(3), 125-134. [2] López-Martínez, C., and Alonso-González, A. (2014). Assessment and estimation of the RVoG model in polarimetric SAR interferometry. IEEE transactions on geoscience and remote sensing, 52(6), 3091-3106. [3] Alonso-Gonzalez, A., and Papathanassiou, K. P. (2018, June). Multibaseline Two Layer Model PolInSAR Ground and Volume Separation. In EUSAR 2018; 12th European Conference on Synthetic Aperture Radar (pp. 1-5). VDE.
10:10 -
Towards The Estimation Of Soil Parameters Under Agricultural Vegetation By Means Of PolInSAR Meassurements
Hecht, Emanuel - German Aerospace Center (DLR), Germany
The potential of soil moisture inversion based on polarimetric SAR interferometry (PolInSAR) observables is widely acknowledged, even when the soil is covered by the canopy of various, agricultural crops. However, a robust, model-based algorithm for reliable soil moisture estimations hasn't been established yet. Over vegetated fields, soil moisture estimations rely upon the separation of the volume and ground scattering contributions. The lack of detailed knowledge of the plant’s dielectric and structural characteristics makes it difficult to use complex scattering models for the interpretation and inversions of the observations. Simple, single or multiple layer models, like the random-volume-over-ground model (RVoG), can be used to estimate the ground and volume powers in a maximum likelihood sense, leaving us with an ill-posed problem. In the single baseline case, this manifests in the ambiguous estimation of the volume coherence. The ill-posedness, however, holds regardless of the number of used polarimetric channels or baselines and thus regularization is implicitly needed. Usually, this is done by making assumptions on the vertical structure of the volume, i. e. the crop canopy, or, more generally, on the polarimetric properties of different layers. Much work went recently in the assessment of fitting the RVoG-model to the coherence region of the PolInSAR coherency matrix, but uncertainties remain concerning the question how regularization methods impact the performance of inversion attempts. In this work, we investigate the integrity of possible choices of the volume coherence in relation to underlying surface and vegetation characteristics. Starting from the PolInSAR coherence region, the impact of different regularizations on the estimated ground parameters is examined for various volume profiles. Based on that, options for an adaptive profile parametrisation toward improved soil parameter inversion are analysed and discussed. The different approaches are implemented and applied on on airborne F-SAR data at different frequencies from the CROPEX 2014 campaign over agricultural fields near Wallafing, Germany. The obtained results are compared against the available reference data.
10:30 - 11:00
11:00 - 12:20
11:00 -
PolSARpro-Bio Edition : The new ESA toolbox for ESA & third party fully PolSAR missions
POTTIER, Eric (1); SARTI, Francesco (2); FITRZYK, Magdalena (3); PATRUNO, Jolanda (4) - 1: University of Rennes 1 / IETR, France; 2: ESA-ESRIN; 3: RSAC c/o ESA-ESRIN; 4: RHEA System S.A. c/o ESA-ESRIN
The objective of this paper is to make a general presentation of the future version of the ESA PolSARpro software toolbox which is now named : PolSARpro-Bio Edition software. In order to prepare for the scientific exploitation of BIOMASS and SAOCOM missions, ESA wishes to develop a new toolbox version called PolSARpro-Bio Edition adding appropriate and state-of-the-art algorithms for advanced processing of polarimetric data (including polarimetry, Pol-InSAR and tomography) and new readers and functionality while improving the user support elements and performance. PolSARpro-Bio Edition will pursue the philosophy of ESA Toolboxes by offering an open source scientific toolbox readily available to exploit data from the first fully polarimetric ESA missions (BIOMASS) and new (like the SAOCOM or the GF-3 Chinese missions) or upcoming third party missions. In addition PolSARpro-Bio Edition will also provide the new generation students with basic and advanced training for the scientific exploitation of polarimetric SAR data and showcase some applications where fully polarimetric data brings unique benefits. For this, the PolSARpro project provides educational software that offers a tool for self-education in the field of Polarimetric SAR data analysis and a comprehensive suite of functions for the scientific exploitation of fully and partially polarimetric data sets. The PolSARpro software establishes a foundation for the exploitation of polarimetric techniques for scientific developments and stimulates research and applications developments using Pol-SAR (Polarimetric SAR), Pol-InSAR (Polarimetric – Interferometric SAR), Pol-TomoSAR (Polarimetric – Tomographic SAR) and Pol-TimeSAR (Polarimetric – Time Series SAR) data. The PolSARpro-Bio Edition activity falls under action line 2 and 1 of the SEOM program (Scientific Exploitation of Operational Missions) which is a new program element within the fourth period (2013-2017) of ESA Earth Ob-servation Envelope Programme (EOEP-4). The PolSARpro-Bio Edition Software will be released in January 2019 and will be avalaible for free download. A global overview of the objectives and functionalities of the PolSARpro-Bio Edition Software, the new scientific toolbox for ESA & third party fully polarimetric SAR missions, will be presented during the POLINSAR 2019 Workshop.
11:20 -
Ambiguities in Poincare sphere orbit signatures of elementary scatterers under orthogonal transformations of the HV basis for dual-pol mode acquisition
Ratha, Debanshu (1); Rao, Y. S. (1); Pottier, Eric (2) - 1: Indian Institute of Technology, Bombay, India; 2: University of Rennes 1, France
European Space Agency’s (ESA) Sentinel-1 mission provides a comprehensive coverage of earth in dual polarization mode (VV+VH) with a frequent revisit time of six days from a constellation of two satellites. The transmission occurs in vertical polarisation (V) only, hence the received backscatter can be analyzed as a Stokes vector on the Poincare sphere. Backscatter from targets is affected by azimuthal orientation and slope of the terrain. Hence, many-a-times an orthogonal transformation of HV basis is applied to identify targets. This orthogonal transformation of HV basis is brought about by three angle parameters i.e. phi, tau, and alpha. Of the three parameters, the phi and the tau are responsible for the shape of the polarization ellipse formed by the backscatter. In this work, we analyze the behavior of elementary scatterers in (VV+VH) mode under the orthogonal transformation of HV basis. The orbits of the response of the elementary scatterers are traced over the Poincare sphere: first by varying the phi and tau parameters independently and then simultaneously in a later case. We observe interesting orbit patterns for the elementary scatterers some of which overlap. Thus, the elementary scatterers cluster into natural groups based on these orbit patterns. This leads to ambiguity in identifying elementary scatterers in dual-pol mode. Thus through this study, we identify the ambiguous and non-ambiguous orbit signatures corresponding to elementary scatterers available in the literature. We also introspect on some of the strategies that may help us deal with these ambiguities when going for model-based applications using dual-pol data.
11:40 -
Quantitative Evaluation of Ionospheric Distortions on Spaceborne L-band SAR under Faraday rotation
Kim, Jun Su; Papathanassiou, Kostas - DLR, Germany
For low-frequency spaceborne SAR configurations, the distortions induced by the by the ionosphere are of prime interest as they critically impact the estimation accuracy of bio- and geophysical parameters from the SAR data. The majority of the ionosphere-induced distortions can be compensated with a sufficiently accurate TEC or differential TEC knowledge that can be derived in many cases - directly or indirectly - from the SAR data itself. This is for example the case for ionospheric phase distortions in interferometric measurements that can be compensated by the use of split spectrum approaches. However, the Faraday Rotation distortion of the scattering matrix induced by the rotation of the polarisation plane of the transmitted and scattered pulses w.r.t. the propagation vector in the linear basis during the propagation through the ionosphere is only evaluated from quad-pol data. Even if the amount of FR is known, through the knowledge of TEC, the correction of FR is possible only when the quad-pol measurements are available. In this paper we assess the importance of polarimetric measurements for the compensation of Faraday rotation distortion by quantifying and evaluating the impact of Faraday rotation at L-band measurements for a wide range of applications. Using 20 years of global TEC measurements from the GNSS network we investigate Faraday rotation induced by the background ionosphere for a Sentinel-like orbit. We do this for different natural scatterers and landcover types at different latitudes. Cumulative probability functions of the distortions are constructed in order to evaluate the level of ionospheric effect on SAR polarimetry. In a second step, depending on the acquisition scenario, different “non-quad-polarimetric” corrections approaches are implemented and compared against the quad-polarimetric correction performance. This way a quantification of the Faraday rotation impact at L-band and at the same time the importance of quad-polarimetric measurements is assessed.
12:00 -
Multidimensional distance geometry analysis for classification of PolSAR images in Arctic scenarios
Marinoni, Andrea; Johansson, Malin; Espeseth, Martine; Brekke, Camilla - UiT - the Arctic University of Norway, Norway
In this paper, we present a new method for the classification of complex Polarimetric synthetic aperture radar (PolSAR) datasets. We focus our attention on images acquired over the Arctic region, since atmospheric and water conditions can induce strong nonlinear effects on the scattering and polarimetric attributes. We first reformulate the classification problem of PolSAR images as an unmixing issue, i.e., we aim to produce maps of areal percentage (i.e., abundance) of the different classes. The investigation of PolSAR datasets is expressed in terms of distance geometry, where all properties are expressed in terms of pairwise distances. We derive distance-based equations for the calculation of multidimensional volumes and for solving the abundance estimation problem, while obeying some of the constraints on the abundances. Next, we introduce geodesic distances, defined as shortest-path distances along a nearest-neighbor graph in the data set. Under certain curvature conditions of the data manifold, we can use these geodesic distances in the distance-based exploration algorithm so that the classification is performed taking the structure of the data manifold into account. This yields an algorithm that is capable of nonlinear characterization, and has a much better performance than the computationally intensive solution where some linear classification algorithm would be simply preceded by a nonlinear dimensionality reduction. Experimental results on datasets acquired for sea ice and oil spill classification will strengthen our approach. Recent advances in imaging hardware for remote sensing have enabled images with high spectral, spatial and temporal resolution. Further, high-performance computing strategies and architectures have made it possible to develop brand new algorithms and methodologies for thorough analysis of large amounts of diverse remotely sensed data. These developments and achievements paved the way to the use of reliable Earth Observation (EO) data to gather relevant information in several research fields, such as sea ice classification and water pollution study. Specifically, thorough description of physical-chemical composition of solid and aerial composites, change detection among large EO temporal series, and segmentation of remotely sensed imagery for clustering and feature selection play a key-role in disciplines that aim at investigating and exploring human-environment interactions. PolSAR has been widely used as an important tool for remote sensing applications. The ability to operate without being affected by lack of day light, nor weather conditions, is essential for Earth science applications in an Arctic environment. PolSAR systems are also capable of providing high spatial image resolution, i.e., accurate description of the physical phenomena occurring on the surface of the Earth. Several techniques have been recently proposed to analyze polarimetric radar data for environmental and sustainability development applications. Several features have been derived from the received wave scattering to be used as indicators of the target properties, enabling discriminate surface types and classes. The resulting information can be used to generate images that optimize the return from different surfaces, to map land cover and also to gain a better understanding of the radar wave and scattering medium interaction. For such remote sensing tools, speckle interference pattern appears in the form of a positive definite Hermitian matrix, which requires specialized models and makes change detection, classification and segmentation challenging. In fact, several methods have been proposed in technical literature to properly address investigation of covariance and coherence features for classification, and thorough characterization of the targets. Hence, architectures based on statistical approach have been developed and studied in order to achieve a precise outline of biophysical properties from the proper combination of polarimetric features and texture analysis outcomes. However, statistical-based classification might not be adequate for describing images that are very complex from a geometrical and polarimetric point of view. Examples include secondary and higher order reflections, shallow and open water environments, intricate mineral mixtures, etc. Characterization and interpretation of such scenarios have often been handled by extensively modeling the source of the nonlinear effects (e.g., multiple scattering), or by employing more model-independent methods for dealing with nonlinearity (e.g., kernel-based processing and artificial neural networks) based on the multiple features extracted by polarimetric analysis. When the polarimetric features associated with the different classes are non-uniform and their statistical distributions are partially overlapping, these methods can suffer from strong performance degradation, leading to inaccuracy in estimates. An alternative strategy for dealing with nonlinearities in multidimensional data sets is performing a nonlinear dimensionality reduction, yielding a linear space of reduced dimensionality, followed by traditional algorithms based on the assumption of linear combination of polarimetric features from the different classes. Most nonlinear dimensionality reduction algorithms are data-driven and unsupervised, and use a geometrically oriented approach based on manifold learning. Typically they consist of a mapping that preserves some global or local relationship from a high-dimensional manifold (constituted by the source data) while projecting it to a lower dimensional linear space. The subsequent linear operations may relate to any of the conventional classification or compression techniques often performed on PolSAR data and attributes, yielding a two-step process for coping with nonlinear data sets. An important disadvantage of most nonlinear dimensionality reduction techniques is their high computational cost and memory requirements, making them rather impractical for use when dealing with sizable PolSAR scenes. The scalability might be achieved by aligning parallel executions of preprocessing steps (such as ISOMAP) on image tiles, while streamlining the computation of geodesic distances and definition of local neighborhoods on the manifold. Other methods resort to explicit use of supervision, i.e., by allowing the manifold structure to be used as input in offline learning of a classification model. Despite these improvements, it can be observed that nonlinear dimensionality reduction of large polarimetric data is usually only proposed for applications where sizable nonlinear effects can be expected beforehand. Arctic sea ice characterization and ocean modelling can be considered as such examples, highlighting the value of the method presented in this paper for these applications
12:20 - 13:20
13:20 - 14:40
13:20 -
Detection of wind turbines using multi-polarization C- and X-band spaceborne SAR data
Ferrentino, Emanuele (1); Marino, Armando (2); Nunziata, Ferdinando (1); Migliaccio, Maurizio (1); Li, Xiaoming (3) - 1: Università di Napoli Parthenope, Italy; 2: The University of Stirling, Scotland (UK); 3: Chinese Academy of Sciences, China
Wind is a sustainable and alternative resource for producing energy and it has a good reputation of being a green form of electricity. Within this context, wind turbines are widely used at onshore and offshore sites to convert the energy of moving air into electrical power. For this reason, wind turbines are a critical infrastructure whose monitoring is an important issue for both economy and environment protection. Within this context, remote sensing can be an important tool to guarantee an effective and relatively cheaper monitoring. Optical images have the great advantage of being simple to interpret and they are easily obtainable. However, optical radiation is severely affected by cloud cover, solar illumination, and other adverse meteorological conditions. These problems can be solved using radar sensors, which guarantee all-day and almost all-weather acquisitions, together with a wide area coverage. In particular, the Synthetic Aperture Radar (SAR) can be very useful for intertidal zone monitoring purposes, because of its fine spatial resolution. The main goals of this study are to develop multi-polarimetric methods to detect wind turbines into a challenging scenario (i.e. mudflats) using full- and dual-polarimetric SAR data. For this purpose, the detection of wind turbines is undertaken according to the Polarimetric Notch Filter and the change detection approach proposed in [1] and [2], respectively. Experiments, undertaken on actual SAR data collected over the intertidal zone near Jiangsu, China, by the C-band RadarSAT-2 and X-band TerraSAR-X missions show that the proposed methodologies, well detect the wind turbines inside mud flat areas. Furthermore, a detailed analysis shows that polarimetric information always guarantees performance better than the single–polarization counterpart. [1] A. Marino, (2013), “A Notch Filter for Ship Detection With Polarimetric SAR Data", IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 6(3), 1219-1232. [2] A. Marino; S.R. Cloude, and J. M. Lopez-Sanchez, (2013), “A New Polarimetric Change Detector in Radar Imagery”, IEEE Transactions on Geoscience and Remote Sensing, 51(5), 2986 -3000.
13:40 -
Towards a Physical Quantitative Assessment of Model-Based PolSAR Decompositions
Ballester-Berman, J. David (1); Ainsworth, Thomas L. (2); Lopez-Sanchez, Juan M. (1) - 1: University of Alicante, Spain; 2: Remote Sensing Division (NRL), USA
As we thoroughly motivated in [1] the performance of model-based decomposition approaches for parameter retrieval is still an open issue in PolSAR field. Very few studies have hitherto performed an in-depth analysis of the incoherent model-based decomposition concept (i.e. Freeman-Durden concept) for quantitative remote sensing applications. Some noteworthy examples are the works by Jagdhüber et al. [2], Huang et al. [3], Di Martino et al. [4], and He et al. [5] focused on soil moisture inversion. In [1] we emphasized the necessity of analysing the estimation accuracy of the whole set of parameters involved in the physical model. To do so, we took as starting point the progress made by several previous contributions based on the Freeman-Durden approach. More particularly, we focused on the general model proposed by Chen et al. [6] as we consider it includes all the previous improvements reported in the literature concerning the topic. Nevertheless, the purpose of that work could be also accomplished by using any of the other published models (see [7-10] to name a few). We showed that a reasonably overall accuracy can be achieved by including several improvements throughout the inversion procedure. However, only some particular cases were considered and the whole range of input values and the different combinations among them were not employed for such purpose. We claim that there still remain some open issues regarding the quantitative assessment of model-based PolSAR decomposition as also Ainsworth et al. recently discussed in [11]. The performance analysis for real data is hampered by the usual unavailability of the corresponding ground-truth data for comparison. In addition, due to modelling issues there appear some parameters whose validation and interpretation are subject to a high ambiguity, as it is the case of vegetation orientation and randomness. The questions that remain to be answered are: What is the actual role of them? Can we assign them a consistent physical interpretation or are they acting just as fitting parameters? The methodology employed in the present work consists on simulating the coherency matrix according to the general model proposed by Chen [6]. Three main scattering mechanisms are assumed (helix component is neglected here). Our hypothesis assumes that an increase of the entropy would lead to a decrease of the parameter estimation accuracy. This methodology is based on the following steps: 1) Generation of noisy samples by applying Lee's method [12]; 2) Inversion of the whole set of parameters according to the procedure shown in Xie et al. [1] (code available [13]); and 3) Computation of histograms of output parameters, standard deviation and bias. Simulations are carried out for different entropy scenarios. The analysis reveals that even the backscattering powers associated with all three basic scattering mechanisms are estimated with an error higher than 10% for usual scattering scenarios. Low entropy cases where surface dominates, and a double-bounce mechanism exists as a secondary return, lead to an accurate estimation of the whole set of parameters. However, an increasing orientation angle of the dominant surface scattering induces a high error in several output parameters and also in the recovered backscattering powers. This means that parameter retrieval algorithms, but also classification schemes, based on this decomposition values could be compromised for real applications. [1] Xie, Q.H.; Ballester-Berman, J.D.; Lopez-Sanchez, J.M.; Zhu, J. J.; Wang, C.C. Quantitative analysis of polarimetric model-based decomposition methods. Remote Sens. 2016, 8, doi:10.3390/rs8120977. [2] Jagdhuber, T.; Hajnsek, I.; Papathanassiou, K.P. An iterative generalized hybrid decomposition for soil moisture retrieval under vegetation cover using fully polarimetric SAR. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens. 2015, 8, 3911–3922. [3] Huang, X.D.; Wang, J.F.; Shang, J.L., An Integrated Surface Parameter Inversion Scheme Over Agricultural Fields at Early Growing Stages by Means of C-Band Polarimetric RADARSAT-2 Imagery. IEEE Trans. Geosci. Remote Sens., vol. 54, no. 5, pp. 2510–2528, May 2016. [4] Di Martino, G.; Iodice, A.; Natale, A.; Riccio, D. Polarimetric two-scale two-component model for the retrieval of soil moisture under moderate vegetation via L-band SAR data. IEEE Trans. Geosci. Remote Sens. 2016, 54, 2470–2491. [5] He, L.; Panciera, R.; Tanase, M.A.; Walker, J.P.; Qin, Q. Soil moisture retrieval in agricultural fields using adaptive model-based polarimetric decomposition of SAR data. IEEE Trans. Geosci. Remote Sens. 2016, 54, 4445–4460. [6] Chen, S. W.; Wang, X. W.; Xiao, S. P., and Sato, M., “General Polarimetric Model-Based Decomposition for Coherency Matrix,” IEEE Trans. Geosci. Remote Sens., vol. 52, no. 3, pp. 1843–1855, Mar. 2014. [7] Yamaguchi, Y.; Moriyama, T.; Ishido, M.; Yamada, H. Four-component scattering model for polarimetric SAR image decomposition. IEEE Trans. Geosci. Remote Sens. 2005, 43, 1699–1706. [8] Cui, Y.; Yamaguchi, Y.; Yang, J.; Park, S.E.; Kobayashi, H.; Singh, G. Three-component power decomposition for polarimetric SAR data based on adaptive volume scatter modeling. Remote Sens. 2012, 4, 1559–1572. [9] Singh, G.; Yamaguchi, Y.; Park, S.E. General four-component scattering power decomposition with unitary transformation of coherency matrix. IEEE Trans. Geosci. Remote Sens. 2013, 51, 3014–3022. [10] Lee, J.S.; Ainsworth, T.L.; Wang, Y. Generalized Polarimetric model-based decompositions using incoherent scattering models. IEEE Trans. Geosci. Remote Sens. 2014, 52, 2474–2491. [11] Ainsworth, T.L.; Lee, J.-S.; Wang, Y. Model-Based PolSAR Decompositions: Virtues and Vices. In Proceedings of 12th European Conference on Synthetic Aperture Radar (EUSAR2016), Aachen, Germany, 4–7 June 2018; 8676–8679. [12] Lee, J.S.; Pottier, E. Polarimetric Radar Imaging: From Basics to Applications; CRC Press: Boca Raton, FL, USA, 2009 [13] https://personal.ua.es/en/davidb/soil-moisture-variations-in-deforested-tropical-areas-preliminary-results.html
14:00 -
A Novel Scattering Vector Parameterization Method
Yin, Junjun (1); Yang, Jian (2); Joerg, Hannah (3) - 1: University of Science and Technology Beijing, Beijing, China; 2: Tsinghua University, Beijing, China; 3: German Aerospace Center – Microwaves and Radar Institute, Wessing, Germany
In this study, we propose a new scattering vector parameterization method for interpreting coherent scattering. This new scattering model, which has the same formation as the alpha-beta model, consists of 5 polarimetric parameters including the normalization factor. Experimental results show the new scattering model is capable of distinguishing forest from the oriented urban areas when only polarimetric parameters are considered.
14:20 -
Monte Carlo Simulation study of stochastic distances applied on k-means algorithm for Fully Polarimetric SAR images
Carvalho, Naiallen; Sant'Anna, Sidnei; Bins, Leonardo - National Institute for Space Research - INPE, Brazil
Monitoring the processes of Earth surface, like deforestation, urban growth, and natural phenomena are essential for maintenance of the ecosystems. Wherefore the different image formats availability, from optical to microwaves frequencies, is increasingly needed. Those images classification is an important tool to comprehend the process that changes the Earth terrain in the present allowing us to predict the future. Different classification strategies have been used to extract information from Fully Polarimetric Synthetic Aperture Radar (PolSAR) images, among of these, the unsupervised methods have the advantage of no needed of prior information, as labeled data. Along with that, the PolSAR images have the advantage of allowing the scattering matrix generation, from which moisture, surface roughness, shape, and geometry information could be extracted. But, as commonly occur with remote sensing data, the images are an earth terrain representation compilation and some regions have a heterogeneous and confusing spectral information and the non-negligible speckle noise, this scenario makes the use of statistical approach a powerful allied in the classes definition. Since PolSAR images could be represented as a n-look covariance matrix, obtained from the scattering matrix, it follows a multivariate Wishart Distribution. Based on this distribution, in this work, we study the unsupervised classification method K-means using a statistical approach. The idea is to represent the PolSAR images as covariance matrices in a feature space, and perform the data clustering based on the minimal stochastic distance between two Wishart Distribution, instead of using the traditional Euclidean distance. We considered five stochastic distance: Bhatacharyya, Kullback-Leibler, Rényi, Hellinger, and Chi-Square. And we considered two cases of study: the first one using simulated images, and as a validation step for this case, a Monte Carlo simulation to analyze the possible results of each distance was performed. The second case using a cut over part of the Tapajós National Forest and surrounding area. This area is considered an important conservation unit in the Brazilian Amazon. The PolSAR images were obtained by the satellite PALSAR 1. In order to compute the classification quality, samples over the image were collected and based on it the accuracy was computed by the kappa coefficient. For each set of PolSAR image six version of k-means was executed, five using the stochastic distances and the sixth using the traditional Euclidian distance, the results showed that the statistic approach achieves higher accuracy than the traditional one, and over the five stochastic distance Bhatacharyya had the best evaluation.
14:40 - 15:00
15:00 - 15:20
15:00 -
MULTI-TEMPORAL QUAD-POLARIMETRIC CHANGE MATRIX FOR AGRICULTURAL FIELDS MONITORING
Silva, Cristian (1); Marino, Armando (1); Lopez-Sanchez, Juan Manuel (2); Cameron, Iain (3) - 1: University of Stirling, United Kingdom; 2: University of Alicante, IUII, Alicante, Spain; 3: Environment systems Ltd, Aberystwyth, United Kingdom
In this work, a new way to extract temporal polarimetric information from a stack of co-registered images is presented. This method considers not only polarimetric evolution of consecutive acquisitions but also includes polarimetric changes between every image with respect to the rest of images in the stack. The methodology is tested for different crop types exploiting C-band quad-polarimetric RADARSAT-2 data over rice fields in Seville, South-West of Spain and the Indian Head in Canada as part of the Agrisar 2009 campaign.
15:20 - 15:50
15:50 - 16:50
15:50 -
EVALUATING THE CLOUDE-POTTIER DECOMPOSITION FOR CROP CLASSIFICATION USING MULTI-TEMPORAL RADARSAT-2 DATA
Ustuner, Mustafa
Ustuner, Mustafa (1); Balik Sanli, Fusun (1); Abdikan, Saygin (2); Erten, Esra (3,4); López-Martínez, Carlos (5) - 1: Department of Geomatic Engineering,Yildiz Technical University, Istanbul, Turkey; 2: Department of Geomatics Engineering, Bulent Ecevit University, Zonguldak, Turkey; 3: Faculty of Science, Technology, Engineering & Mathematics, The Open University, Milton Keynes, United Kingdom; 4: Department of Geomatics Engineering, Istanbul Technical University, Istanbul, Turkey; 5: Remote Sensing and Natural Resources Modelling Group, Luxembourg Institute of Science and Technology, Luxembourg
Crops have very dynamic structure and various scattering characteristics over the phenological growth stages. Polarimetric SAR (PolSAR) images are the important data source for large-scale/time-critical agricultural practices and able to provide the details about the structural and geometrical properties of the crops. In this study, the potential of the Cloude-Pottier decomposition for the crop classification using multi-temporal Radarsat-2 SAR (single look complex full polarimetric data with fine quad-polarization acquisition mode) data is evaluated. Cloude-Pottier decomposition (also known H-A-α decomposition) which is based on the eigen-decomposition of the coherency matrix [T] compute three polarimetric parameters: entropy (H), anisotropy (A), mean alpha angle (α). This mean alpha angle (α) is the weighted average of alpha1 (α1), alpha2 (α2) and alpha3 (α3) angles. The alpha angle (αi) can sometimes be preferred over mean alpha angle since mean alpha angle can be very noisy especially for low Entropy scatterers. To investigate the impact of each alpha angle (αi) as well as other polarimetric parameters of Cloude-Pottier decomposition for the crop classification, five different dataset were generated: (I) H-A-α1, (II) H-A-α2, (III) H-A-α3, (IV) H-A-α, (V) Lambda1-Lambda2-Lambda3 (eigenvalues). For the classification of the crop types (maize, potato, summer wheat, sunflower, and alfalfa) in the test site, two different classification models (random forest and extremely randomized trees) were implemented. The classification of the PolSAR data was exploited in the radar geometry. The experimental results demonstrate that dataset-II obtained higher classification accuracy (79.59% for random forest and 79.85% for extremely randomized trees) than other datasets except dataset-V for both classification models. The highest classification accuracy was obtained from dataset-V as 88.23% for extremely randomized trees while the lowest accuracy was received by dataset-III as 74.35% for random forest. Dataset-I also outperformed dataset-IV in terms of classification accuracy. This experimental study suggests that alpha angles (i.e. alpha2 or alpha1) can be considered as an alternative parameter instead of mean alpha angle as the third component of Cloude-Pottier decomposition for the crop classification. In addition, the findings prove the multi-temporal PolSAR data, by itself, can be sufficient for crop classification with satisfactory results.
16:10 -
Crop Growth Monitoring Using Compact-Polarimetry SAR Data
Hosseini, Mehdi (1); McNairn, Heather (2); Mitchell, Scott (1); Davidson, Andrew (2); Dingle Robertson, Laura (2) - 1: Carleton University, Ottawa, Canada; 2: Agriculture and Agri-Food Canada
Total above ground biomass and Leaf Area Index (LAI) are two important crop biophysical parameters linked to crop yields (Lobell, 2013) and as such, can be useful surrogates for yield potential. Many studies have demonstrated that optical satellite data are able to estimate biomass and LAI of various crops (Günlü et al., 2014; Hosseini et al., 2015; Kross et al., 2015). However, cloud cover impedes visible-infrared wavelengths making it particularly challenging to rely exclusively on this class of satellites for monitoring rapidly developing crops. Conversely longer microwaves used by Synthetic Aperture Radars (SARs) permit collection of data even during cloudy conditions. Recent studies have demonstrated that SAR sensors are also able to estimate biomass and LAI (Betbeder et al., 2016; Hamdan et al., 2011; Hosseini and McNairn, 2017). SAR scattering changes as crop structure develops, with backscatter also impacted by soil moisture. In this study, simulated compact polarimetry Synthetic Aperture Radar (SAR) satellite data are assessed for estimating the biomass and LAI for four Canadian important crop types; corn, soybeans, wheat, and canola. The compact polarimetry mode provides broader swath width which is important for regional and global monitoring. The objective of this research is to make the users ready for the RADARSAT Constellation Mission (RCM) which was planned to be launched in February 2019. The RCM configuration is a transmit right circular polarization with horizontal and vertical linear polarizations received. In 2012, the Soil Moisture Active Passive (SMAP) Validation Experiment (SMAPVEX12) field campaign was conducted in Carman, Manitoba, Canada (McNairn et al., 2015). Multiple crops and soil parameters including LAI, above ground biomass, crop height, soil moisture, and surface roughness were measured in 55 agricultural fields. This data along with their corresponding quad polarimetric RADARSAT-2 imagery is used for our modeling and the validations. Simulated compact polarimetric data are produced using the Sentinel Application Platform (SNAP) which is jointly developed by Canadian and European companies (http://step.esa.int/main/toolboxes/snap/). A total of 43 polarimetric parameters are generated. Time series of these parameters are correlated with crop biophysical parameters. Few of these parameters including the volume-to-surface scattering ratio are strongly correlated with the crop biomass and LAI and demonstrated the potential of compact polarimetry SAR data for crop growth monitoring.
16:30 -
Modelbased Assessment of the Ground Polarimetry in Crops estimated using Polarimetric Interferometric SAR
Joerg, Hannah (1); Alonso-Gonzalez, Alberto (1); Papathanassiou, Konstantinos (1); Hajnsek, Irena (1,2) - 1: German Aerospace Center (DLR); 2: ETH Zurich
Electromagnetic scattering mechanisms in agricultural scenarios, i.e. in dependency of dielectric and geometric soil and plant parameters, are highly complex. Especially towards an inversion of soil moisture under crops from synthetic aperture radar (SAR) measurements, it is necessary to separate the scattering mechanisms occurring on the ground, i.e. surface (soil) and dihedral (soil-trunk interactions) scattering, from the volume scattering in the vegetation layer above. While the observation space provided by polarimetric SAR measurements is limited in this respect, polarimetric interferometric SAR techniques, in a single- or multi-baseline configuration, can be used to separate the scattering mechanisms along height. Utilizing a two-layer model consisting of an isotropic volume layer on top of an impenetrable polarization dependent ground layer can be employed to estimate the volume only interferometric coherence and thus the ground and volume polarimetric covariance matrices [1-3]. However, the estimation of the interferometric volume coherence is intrinsically ambiguous and subject to regularization constraints, independently of the number of available baselines. As a direct consequence, also the estimation of the polarimetric ground covariance matrix is ambiguous impacting its physical interpretation and any parameter inversion attempt [4]. The first part of this work aims to foster the understanding of the feasibility and the limitations for an inversion of soil moisture from such an isolated ground component. It is assumed that the isolated ground component can be modelled as the sum of surface and dihedral scattering and therefore depends on the soil and trunk dielectric constants and the surface-to-dihedral power ratio. Based on this, a comprehensive sensitivity analysis of the model in dependency of incidence angle and scattering scenario will be conducted and possible inversion strategies evaluated. In the second part, the feasibility of the ground model and inversion strategies discussed in the first part will be assessed using ground covariance matrices estimated from experimental SAR data at L-band acquired by DLR’s airborne sensor F-SAR in the frame of the CROPEX 2014 campaign. One important issue is to evaluate the impact of the regularization required to obtain a unique ground estimate in this respect. The availability of fully polarimetric multi-baseline L-band SAR data on different dates further allows assessing scattering scenarios with different vegetation characteristics (i.e. crop type, phenological stage, vegetation water content) and soil moisture levels. [1] S. Tebaldini, “Algebraic Synthesis of Forest Scenarios from Multibaseline PolInSAR Data”, IEEE Trans. Geosci. Remote Sens., vol. 47, no. 12, pp. 4132-4142, Dec. 2009 [2] M. Pardini, and K. Papathanassiou, “On the Estimation of Ground and Volume Polarimetric Covariances in Forest Scenarios with SAR Tomography,” IEEE Geosci. Remote Sens. Lett., vol. 14, no. 10, pp. 1860-1864, Oct. 2017. [3] A. Alonso-Gonzalez, and K. P. Papathanassiou, “Multibaseline Two Layer Model PolInSAR Ground and Volume Separation,” Proc. of EUSAR 2018, vol. 12, Aachen, Germany, 2018. [4] H. Joerg, M. Pardini, A. Alonso-Gonzalez, K. P. Papathanassiou and I. Hajnsek, “Investigating Soil Moisture Dependency of the Ground Polarimetry under Agricultural Vegetation Estimated by SAR Tomography,” Proc. of EUSAR 2018, vol. 12, Aachen, Germany, 2018.
16:50 - 17:10
17:10 - 19:10
Extending parameter space analysis of model based decomposition for coastal wetland classification
kwok, expo
In view of the fact that the results of scattering model-based decomposition is relatively simple and the deep scattering mechanism has less exploration, this paper defined two class parameters: model-entropy and model-anisotropy based on model-based decomposition, by borrowing ideas from the information theory based on the well-known Cloude decomposition. because this type of decomposition result is known to be the dominant scattering mechanism, it is not necessary to define the scattering angle parameters of the decomposition.according to the physical meaning of the two types of parameters, we can define multiple parameter combinations that highlight different scattering information, so as to expand the range of quantitative analysis of its scattering mechanism, and use a random forest algorithm that can select and contribute to a variety of features to perform classification experiments. In this paper, real polarimetric SAR and simulated polarimetric SAR data are used as research data. Real SAR data includes China's first multi-polarization SAR satellite GaoFen-3 whose highest resolution can be 1m and AIRSAR data. and China's largest reed wetland, the Liaohe Estuary National Nature Reserve, is used as research area. There are large areas of reeds in this area, special plant Suaeda salsa and single-season rice in northeast China, and numerous water systems (including seawater, reservoir ponds, and river waters). According to the interpretation of two class parameters, it shows that this method greatly enriches the scattering space range based on model decomposition, and can be applied to the decomposition algorithm analysis of different number of components. It can significantly distinguish the difference between different decomposition algorithms and improve the polarization target classification accuracy. In addition, it provides an effective idea that quantitative evaluation of this type of decomposition.
The distinction between thawed/frozen soils in Alaska by multi-temporal Sentinel 1 radar data
Rodionova, Natalia
This paper considers the question of determining frozen/thawed status of 5 cm upper soil layer for seven ground stations in Alaska with latitude from 650 to 700 N by using radar Sentinel 1 C-band data for the period 2016-2017. Determine the status of frozen/thawed soil was carried out in two ways: using only radar data with finding the backscatter coefficient threshold, when the temperature in the upper soil layer falls below 00C, or by using both radar data and ISMN ground-based measurements of soil temperature. In the latter case, the correlation between the backscattering coefficient and the soil temperature measured in upper 5 cm soil layer was calculated. Regression models were developed and radar backscatter thresholds for frozen soil were found. Local frozen/thawed soil maps were created. The comparison of the backscatter coefficient threshold values obtained in two ways showed their closeness, which tends to decrease with decreasing correlation between radar and ground ISMN data. Linear regression model between radar backscatter threshold and the area latitude was developed.
Anisotropic Scattering Detection for Characterizing Polarimetric Circular SAR Multi-Aspect Signatures
Li, Yang; Lin, Yun; Wang, Yanping; Hong, Wen
Azimuth anisotropic scattering signatures provide richer features for synthetic aperture radar (SAR) terrain classification and target recognition. Compared with traditional stripmap SAR mode, circular SAR (CSAR) utilizes the backscattering data of a target over a full 360 degrees azimuth angle. The isotropic scattering of media can be added coherently to achieve a high resolution image. However, CSAR subaperture images also contain anisotropic scattering information, allowing differentiation of targets of similar structure from other classes. CSAR subaperture combination is usually processed by coherently or incoherently adding. In these processes, the distinctive scattering signatures in each aspect are mixed together and become an average feature, which is difficult to separate, causing a loss of this information. In this paper, we propose a polarimetric CSAR anisotropic scattering detection framework to characterize multi-aspect and fully polarimetric SAR signatures of point-like and distributed targets. We applied this framework to quantify and rank media polarimetric scattering dissimilarity over all aspects and to determine whether the most different one shows anisotropy by use of constant false alarm rate (CFAR) detection. Furthermore, we demonstrated the monotonicity of CFAR detection function and incorporated this function to decrease the complexity of the anisotropic scattering test. To test this approach, we applied the multi-aspect series generated by this algorithm to analyze the anisotropic scattering effects using a set of airborne P-band fully polarimetric circular SAR data acquired by the Institute of Electronics, Chinese Academy of Science (IECAS). The results indicate the framework can retain anisotropic scattering and extract a series of new multi-aspect polarimetric SAR signatures for terrain classification.
A MACHINE LEARNING APPROACH FOR ACCURATE CROP TYPE MAPPING USING COMBINED POLARIMETRIC SAR AND OPTICAL TIME SERIES DATA
Tufail, Rahat; Javed, Muhammad Asif; Ahmad, Sajid Rashid - College of Earth and Environmental Sciences, University of the Punjab, Pakistan
A country's food needs primarily rely on its own agriculture resources and requires reliable information related to crop health, their distribution and acreage estimation to manage and monitor resources for implementation of a sustainable agricultural system. Different methodologies have been used to collect this information but the availability of earth resource satellites data with increasing spatial, spectral and temporal resolutions such as the European Space Agency (ESA) Copernicus programme's satellites Sentinel-1 and Sentinel-2 creating more practicability to generate crop type maps. Sentinel-1 and Sentinel-2 both operating in a constellation of a twin satellites carries a C band Synthetic Aperture Radar and Multispectral Instrument (MSI) respectively. Vertical transmit and horizontal receive (VH) and vertical transmit and vertical receive (VV) channels used to exploit the temporal backscatter of crops present in the study area. In this research, a machine learning random forest classification algorithm run for accurate crop type mapping for the combination of SAR (Sentinel-1) and optical (Sentinel-2) time series data. Random Forest classifier produced a considerable increase in accuracy of crop type mapping as it uses ensemble decision trees trained on sample data which permit the vote in favor of the most popular class. Main objectives of this study were to investigate the classification accuracy for different data combinations. Three plots of data tested i) Sentinel-1 ii) Sentinel-2 iii) Sentinel-1 & Sentinel-2. Combination of SAR and optical data turns out with a best overall accuracy of 97 %, and a kappa coefficient of 0.97. Spaceborne SAR and optical data add a new aspect of crop type mapping which consequently increases the classification accuracy by including valuable parameters and beating the drawbacks of each other. The results also emphasized the classification of time series data in agricultural mapping than classifying a single date image. By Comparing the results, it can be concluded that by combining all-weather accessible SAR and spectrally rich data accomplished more accurate outcomes and it would be an imperative advance for the future endeavor to estimate crop biomass and biophysical parameters.
Forest biophysical parameters retrieval using L-band airborne multi-baseline UAVSAR datasets
Awasthi, Shubham; Jain, Kamal - Indian Institute of Technology Roorkee, Uttarakhand, India
Forest are one of the important part of ecosystem. They play a crucial role in the carbon cycle. A significant amount of the carbon is stored in the form of the forest biomass. These forest cover areas and their stored carbon stock are affected by various natural and anthropogenic factors. In the recent years, due to the various factors such as forest fires, climate change, and forests encroachments there has been a continuous depletion in the forest cover. Hence, in this present scenario regular monitoring of the forest areas is an urgent need. This study focuses on the utilization of PolInSAR and SAR Tomography technique for the forest biophysical parameters retrieval. UAVSAR datasets of AfriSAR mission operating in L-band (1.5 GHz) frequency region with the azimuth resolution of 1.5m* 12m. The used datasets were acquired on the date 15 June 2015 for the Rabi forest area of Gabon in Africa. Using these datasets the coherence optimization was done and the effect of the spatial baseline and interferometric wave number (Kz) was analyzed on the forest estimation. The three-stage inversion technique was implemented for this PolInSAR based height inversion. The appropriate baseline estimation for the correct forest height estimation was done. In addition, implementation of SAR Tomography based forest height retrieval was also done. The aperture synthesis in the radar cross range direction was able to resolve the forest heights in the vertical direction. The Beamforming and Capon based spectral estimation technique was used for tomographic implementation. The comparative analysis of the PolInSAR band SAR Tomography retrieved forest heights was done. The mean forest height retrieved using these technique was 31m. The DSM of the forest area was generated using airborne Land, Vegetation, and Ice Sensor (LVIS) based airborne LIDAR point cloud datasets which was used for the forest height validation.
A New X-Bragg Scattering-Based Approach for Bare Surface Discrimination
Tahraoui, Sofiane; Ouarzeddine, Mounira
In the last two decades, the use of Synthetic Aperture Radar (SAR) for remote sensing purposes has been significantly developed due to improvements in the quality and in the availability of the images, thanks to SAR polarimetry technique, which has increased further the range of applications of the sensed data. The use of PolSAR allowed the retrieval of geometrical properties and geophysical parameters such as shape, roughness, texture, and moisture content, considerable accuracy. The objective of this paper is to present a new technique to discriminate bare soil from other contribution within the pixel. The procedure is based on the conformity of the canonical form of the X-Bragg coherence matrix. The approach is valid within a limited range of surface roughness. Surface discrimination is performed according to a decision based on a fixed threshold. The methodology was evaluated using simulated and real data and result seem promising.
A PolSAR Scattering Power Factorization Framework using a Geodesic Distance
Ratha, Debanshu (1); Bhattacharya, Avik (1); Frery, Alejandro C. (2); Pottier, Eric (3) - 1: Indian Institute of Technology, Bombay, India; 2: Universidade Federal de Alagoas, Brazil; 3: University of Rennes 1, France
This paper presents a novel scattering power factorization framework in radar polarimetry using a geodesic distance between the 4 x 4 real Kennaugh matrices of the observed and the elementary targets (viz. dihedral, trihedral, dipole etc.). The framework provides both qualitative and quantitative estimates of the dominance of elementary scattering mechanisms in a pixel. It is also flexible in terms of the number of elementary models against which an observed backscattering may be compared. Under this framework, the observed scattering is evaluated in terms of scattering similarities which provide the dominance of scattering mechanisms. This is then further utilized for a convex splitting of unity to obtain the intermediate weights. This leads to the weights being the product of similarity and dissimilarity of the observed pixel with the elementary scattering models. Finally, these weights are modulated with the total power (Span) to obtain the non-negative scattering powers. The results are shown for full polarimetric single-look ALOS-2 L-band dataset and a multi-look RADARSAT-2 C-band dataset.
Atmospheric Effects on Satellite Synthetic Aperture Radar at X-Band and above
Mori, Saverio; Biscarini, Marianna; Marzano, Frank S.; Pierdicca, Nazzareno
Spaceborne synthetic aperture radars (SARs) operating at L-band and above are nowadays a well-established tool for Earth remote sensing in many fields, such as DEM production, monitoring of earthquakes damages, volcanoes, landslides and flooded areas, urban changes detection and many others. Sentinel system follows the long heritage of C-Band SARs matched with near-continuous acquisition, an absolute novelty for SARs. Systems as X-Band, such as COSMO-SkyMed and TerraSAR-X/TanDEM-X are operative for many years, and their second generation is ongoing. New frontiers of research are both technological, exploiting higher frequency SAR and new acquisition modes, and in terms of applications, such as the recent studies on hurricanes monitoring and precipitations observation. In this respect, higher frequency such as Ka-Band are interesting also because the short wavelength allows the implementation on a single platform of single-pass interferometers, both cross-track and along-track, with adequate interferometric sensitivity, baseline independent from location and no temporal decorrelation. Moreover the low penetration in semi-transparent media such as ice, snow, and vegetation Since approximately 2010, a number of European Space Agency (ESA) studies have been done into Ka-band SAR system and mission concepts, and into some of the critical technologies. The CoRe-H2O, finalist for Earth Explorer mission selection, is a Ku and Ka bands SAR platform devoted to cryospheric applications. Another recently proposed mission concept is the KydroSAT one, whose core instrument is a dual Ku and Ka SAR, the first SAR mission specifically proposed for hydrological applications [Mori et al., 2017]. Atmosphere is not transparent at SAR used frequencies, including low ones such as L-Band (e.g. [Melsheimer et al., 1998]). Gases introduce attenuation and path delay in clear-sky conditions. Non-precipitating clouds and ice particles may transmit, scatter or absorb radiant energy that impinges upon them; at Ka-Band their attenuation and path-delay are proportional to water content. Raindrops, due to their oblate shape and dimension, in addition introduce also depolarization. Finally, atmospheric turbulences in the troposphere can induce fluctuations on the refractive-index causing scintillation effects, both in amplitude and phase. These effects, that causes attenuation (up to the complete masking of the scene) and loss of coherence, are usually increasing with frequency, due to the increasing comparability between SAR wavelength and atmospheric particle dimensions. While some of these phenomenons have been analysed and discussed in literature, yet very few studies exist for higher frequencies such as the Ka-Band, moreover expected as the most sensitive. While for most of “traditional” SAR applications atmospheric effects represent a problem to be account for, up to the discarding of affected pixels, this sensitivity to atmospheric particles, in particular to the water based ones represents also a possibility for have new insights on clouds and precipitations. Several studies have exploited these possibilities, demonstrating also the possibility of quantifying precipitations (e.g. [Marzano et al., 2011] and discussing the benefits of a multifrequency approach (e.g. [Marzano et al., 2012]). Within this general context, the ESA contract 4000122671/17/NL/FF/gp "KaSARApp - KA-Band SAR Application Consolidation and Requirement Definition Study" aims to consolidate a Ka-band SAR mission concept, linking user (product-level) observation requirements to mission requirements, and evaluating and highlighting the expected performances for a set of relevant applications. It is the continuation of the ESA contract 4000109477/13/nl/lvh “Ka-Band SAR Backscatter Analysis in Support of Future Applications”; the last was focused on investigating the wave interaction at Ka-band for a widely varying range of targets linked to potential applications in connection to Ka-band InSAR and polarimetry. KaSARapp project will pursue its objectives through the development and use of a sophisticated End-to-End (E2E) performance tool. Atmospheric effects will be taken into account through the forward model of SAR response described in [Marzano et al., 2012] and [Mori et al., 2017]. This model is able to simulate the fully polarimetric response of a SAR system, as function of the operating frequency, incident angle and observed scene. In this respect, ground target can be usually expressed through canonical targets or semi-empirical models (such as [Oh et al., 2002]). Distribution of atmospheric components, such as water-based particles, can be simulated through synthetic distribution, e.g. homogeneous layer of only raindrops or ice particle, with a rectangular shape, unrealistic but more ease to be interpreted. In this respect, the proposed model allows the use of 3-D simulations derived from sophisticated cloud resolving models, such as Weather Research and Forecast (WRF) or System for Atmospheric Modelling (SAM) datasets [Blossey et al., 2007]. These models allow the realistic simulation of volumetric distribution of several water-based particles (such as cloud droplets, raindrops, snowflakes and ice), together with other atmospheric parameter, such as pressure, humidity and temperature, required to model gases, refraction and scintillation effects; moreover resolution that can be very high and suitable for SAR applications (e.g. 250m). Within this project, the forward model of [Mori et al., 2017] has been revised and enhanced in order to include atmospheric gases effects and atmospheric refraction, while the inclusion of atmospheric scintillation effects is ongoing; to our knowledge, each of these effects represent a novelty in the field of SAR forward modelling. In this work, the developed model will be presented. Moreover, within the context of KaSARapp project, it will be presented a preliminary analysis of atmospheric effects on the produced realizations, in terms of attenuation, phase delay, depolarization and scintillation, discussing the actual relevance of the different contribute, and their impact on possible Ka-Band SAR applications.
Effect of Temporal Decorrelation on Polarimetric SAR Interferometric Measurement of Tropical Forest Structure: Results from AfriSAR L- and P- band InSAR Campaign
Ramachandran, Naveen; Saatchi, Sassan; Dikshit, Onkar
PolInSAR measurements for quantifying forest vertical structure and biomass have been designed and implemented in NASA and ESA airborne and satellite missions. From satellite platforms, PolInSAR techniques are performed from two orbital configurations with a non-zero baselines that are often separated in time depending on the repeat-pass designs. These measurements are often impacted by the temporal decorrelation of the interferometric SAR signals due changes of the target characteristics from wind and environmental changes. Here, we explore the effect of temporal decorrelation on retrieving forest vertical structure over Gabonese humid tropical rainforest at L- and P- bands from airborne data collected during the AfriSAR campaign. Our analysis focuses on forest height estimates from conventional and volumetric temporal decorrelation incorporated in Random Volume over Ground and Random Motion over Ground models with fixed and height dependent extinction coefficients. The long-term (few days) and short-term (few minutes) temporal decorrelation have been evaluated using DLR data acquired during AfriSAR 2016 campaign. The model parameters are estimated by minimizing the distance between the predicated and observed coherence in a least square sense. The estimated PolInSAR canopy height using the models mentioned above are compared with NASA’s the Land, Vegetation, and Ice Sensor (LVIS) LIDAR derived canopy heights for quantification of error estimates. The results of the study demonstrate how the effect of temporal decorrelation may introduce uncertainty in forest structure retrieval from the upcoming ESA’s BIOMASS mission and provide potential approaches to reduce the uncertainty.
Sentinel-1 Based Water Level Monitoring for Protected Salt Lake in Turkey
Bilgilioğlu, Burhan Baha (1,2); Erten, Esra (1); Musaoğlu, Nebiye (1) - 1: Faculty of Civil Engineering, Istanbul Technical University, Istanbul 34469, Turkey; 2: Faculty of Engineering and Natural Sciences, Gumushane University, Gumushane, 29000, Turkey
Wetlands are made of up complex ecosystems that provide an ideal habitat for many living species and play an important role in waterway transport and streamflow maintaining. Wetlands are affected by human-induced and natural threats such as groundwater withdrawals, pollution and climate change (e.g. hydrological cycle in wet and dry seasons). All these activities have huge impact on wetlands functions and values, which are highly dependent on water level change. Determination of water level change is an important data source for the conservation and rehabilitation of wetlands. This paper presents evidence of how interferometric SAR (InSAR) techniques with freely available SAR data can be used to monitor wetlands at risk of degradation and loss, using Salt Lake, Turkey, as a case study. The Salt Lake, which has lost 60% of its water over the past 18 years, is the primary source of water in the driest region of Turkey. To monitor its water level fluctuations, a 1-years Sentinel-1 stack composed of 13 dual-pol (VV/VH) images acquired between December 2016 to December 2017 covering dry and wet seasons is employed and InSAR techniques are chosen to assess water level change over the Salt Lake region. In addition, levelling measurements with millimeter level accuracy were made in the time period when the lake was driest in order to compare with InSAR results. Moreover, different polarizations are also investigated to make benefit of maximum interferometric coherence and to determine optimal scattering mechanism in wetlands.
APPLICABILITY OF THE MULTITEMPORAL COHERENCE APPROACH TO SENTINEL-1 FOR THE DETECTION AND DELINEATION OF FOREST FIRES IN THE CONTEXT OF COPERNICUS EMS
Larrañaga, Arantzazu
Larrañaga, Arantzazu (1); de Blas, Teresa (1); Donezar Hoyos, Uxue (1); Ros, Fermín (1); Albizua, Lourdes (1); Steel, Alan (2); Broglia, Marco (2) - 1: Tracasa, Spain; 2: Joint Research Centre (JRC)
The Copernicus Emergency Management Service (EMS) Mapping provides Civil Protection users with accurate and timely geospatial information based on space data combined with other sources during the emergency response cycle. The Copernicus EMS Mapping Validation is the service within the Copernicus Emergency Management Service system in charge of the verification of the results and products created by the different map production services of the Copernicus EMS, Rapid Mapping y Risk and Recovery. The EMS Validation service contributes to the continuous improvement of the service through technical validation of the mapping products, their evaluation based on the user's feedback and the investigation of alternative and innovative technologies. In this framework, the applicability of the MultiTemporal Coherence (MTC) technique using Sentinel-1 data and the software made available by the European Space Agency (ESA), Sentinel Application Platform (SNAP), for the detection and delineation of forest fires was tested. The aim of the study was to follow a previous study in which Sentinel-1 images had been used to delineate fire. For this purpose, four Sentinel-1 images were acquired over an area mainly covered by Mediterranean vegetation that suffered from massive forest fires, together with Shuttle Radar Topography Mission (SRTM) data, and processed using SNAP. The final fire delineation was carried by an Object-Based Image Analysis (OBIA) of the resulting MTC image followed by a visual inspection. The effect of the polarisation, Vertical-Vertical or Vertical-Horizontal as well as the acquisition mode, ascending or descending, were studied in order to assess the differences in the result depending on the input data. The fire delineation derived from Sentinel-1 was compared to the delineation of fire derived using three optical images, a pre-event Sentinel-2 image, and post-event Sentinel-2 and SPOT 6. The first two were used to calculate differences of Burnt Area Index (dBAI) and fire delineation was created by OBIA and photo interpretation with the help of the SPOT 6 image. Data from the European Forest Fire Information System (EFFIS) together with in-situ burnt area delineation provided by the Authorised User of the Copernicus EMS were also used as input data in the refinement of the fire delineation data calculated over optical imagery. Results of the comparison showed the feasibility of using this technique for fire delineation, as well as the recommendations regarding which polarisation gave the best results.
Multi-temporal and multi-frequency analysis to assess forest degradation
Pacheco-Pascagaza, Ana Maria (1); Garcia, Mariano (2); Rodriguez-Veiga, Pedro (1,3); Balzter, Heiko (1,3) - 1: University of Leicester, United Kingdom; 2: Universityof Alcala, Spain; 3: National Centre for Earth Observation (NCEO), United Kingdom
Forest disturbances (i.e. deforestation and degradation) are a serious problem significantly contributing to greenhouse emissions and biodiversity loss [1]. Quantifying the impact of forest degradation on the carbon budget is challenging because of the diversity of definitions, varying scale of the changes, and many drivers that are applying pressure on the forests. As a result, the measuring and mapping of forest degradation is still a technical challenge [1-3]. Radar (SAR) data is very promising for detecting and monitoring forest degradation considering its sensitivity to above ground biomass and forest structure [4]. There is also evidence that the integration of sensors in combination with field data can provide more precise results when monitoring changes in forest [5]. In this research we investigated the potential of integrating multi -frequency Synthetic Aperture Radar (SAR) data: ALOS PalSAR-2, Sentinel-1B and TanDEM-X, in combination with field data to identify and classify variances in structure associated to forest disturbance. Hereafter, we assessed the capabilities full Polarimetric data from ALOS to identify multi-temporal changes in forest structure associated to forest degradation. The study area is located in Bajo Calima a secondary forest in the Pacific coast of Colombia. A wet tropical forest that has been under pressure from gold mining and selective logging at different intensities. Results showed a good approximation to categorize these forests by degrees of disturbance and also indicated that multi-temporal observations Polarimetric ALOS PalSAR 1 and 2 are promising to identify variations in forest structure. References [1] P. Smith, M. Bustamante, H. Ahammad, H. Clark, H. Dong, E. Elsiddig, et al., "Climate Change 2014: Mitigation of Climate Change. Contribution of Work-ing Group III to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change ed O Edenhofer et al," Forestry and Other Land Use, 2014. [2] FAO. (2012b, 11 November 2012). Forestry and climate change: Roles of forests in climate change. Available: http://www.fao.org/forestry/ climatechange/53459/en/ [3] I. D. Thompson, M. R. Guariguata, K. Okabe, C. Bahamondez, R. Nasi, V. Heymell, et al., "An Operational Framework for Defining and Monitoring Forest Degradation," Ecology and Society, vol. 18, p. 20, 2013. [4] A. L. Mitchell, A. Rosenqvist, and B. Mora, "Current remote sensing ap-proaches to monitoring forest degradation in support of countries measure-ment, reporting and verification (MRV) systems for REDD+," Carbon Balance and Management, vol. 12, p. 9, April 17 2017. [5] V. De Sy, M. Herold, F. Achard, G. P. Asner, A. Held, J. Kellndorfer, et al., "Synergies of multiple remote sensing data sources for REDD+ monitoring," Current Opinion in Environmental Sustainability, 2012.
Crop yield estimation by Sentinel-1 radar polarimetric data
Friedl, Zoltán (1,2); Nádor, Gizella (1); Molnár, András (3); Surek, György (4) - 1: Government Office of the Capital City Budapest, Hungary; 2: Eötvös Loránd University, Department of Geophysics and Space Science; 3: Research Institute of Agricultural Economics; 4: MLog Instruments Ltd.
The Synthetic aperture radar is at least as important data for agriculture as optical data. Especially, the radar polarimetric data is connected well with phases of agricultural crop condition. In this presentation the relation of radar polarimetry data and crop yield was examined by using Sentinel-1 dual-pol (VV+VH) time series data including backscatterer coefficients and polarimetric descriptors. The polarimetric descriptors were generated based on H/A/Alpha polarimetric decomposition of covariance matrix. In our study we concentrated on the winter wheat and the sunflower, which are really different type of crops including the growing period, crop features and phenology phases. Sunflower shows a very characteristic temporal profile and it has a higher volume scattering, due to this it is suitable for this analysis. In contrast the winter wheat has a vertical dominance and due to this the volume scattering is lower than double bounce scattering. Thus, crop yield estimation of winter wheat by radar data is complicated more. There were investigated two county in Hungary, which were selected by reason of these are intensive agricultural areas. The reference data concern to 40-50 parcels for each county which were ensured by Research Institute of Agricultural Economics. Our first results are that the correlation between crop yield and the integrated polarimetric descriptors is around 80%. Furthermore, there were analyzed those fields which do not fit into the correlation, and we try to identify reasons of the difference. This is first step to develop a crop yield estimation based on radar data.
Enhanced TomoSAR Imaging through Statistical Regularization
Martín del Campo Becerra, Gustavo Daniel; Reigber, Andreas; Nannini, Matteo
Abstract Synthetic aperture radar (SAR) tomography (TomoSAR) is a powerful remote sensing tool that allows the retrieval of a 3D representation of the illuminated scene [1] - [4]. A set of images, acquired with a different line-of-sight (LOS), is combined coherently using SAR interferometric techniques. Later on, the power spectrum pattern (PSP), in the direction perpendicular to the LOS (PLOS), is recovered using spectral analysis (SA)-based methods. The TomoSAR problem at hand is treated as an ill-conditioned nonlinear inverse problem [5], [6], and is commonly tackled within the direction-of-arrival (DOA) estimation framework [2] - [6]. The DOA-inspired non-parametric techniques, as the conventional matched spatial filter (MSF) and minimum variance distortionless response (MVDR) beamformers [1] - [4], are well suited to cope with distributed targets, since these techniques recover an estimate of the continuous power spectrum pattern (PSP); nonetheless, the achievable resolution highly depends on the span of the tomographic aperture. Alternatively, super-resolved parametric approaches, as multiple signal classification (MUSIC) [3], [4], have the main drawback related to the white noise model assumption that guaranties the separation of the signal and noise sub-spaces. On the other hand, taking advantage of the sparse representations of the cross-range tomographic profiles in the wavelet domain, super-resolved compressed sensing (CS)-based approaches [7], [8], are also employed to solve the TomoSAR inverse problem. However, CS-based techniques often imply a considerable computational burden, due to their iterative nature and due to the non-availability of adapted efficient convex optimization algorithms. To overcome such drawbacks and as an alternative to the aforementioned commonly performed TomoSAR-adapted focusing techniques, statistical regularization approaches can be applied instead, in the context of the statistical decision-making theory. Assuming no a priori knowledge about the statistical distribution of the desired PSP, to be retrieved, and imposing no constrain on linearity, the Bayes minimum risk (BMR) methodology is extended to the maximum-likelihood (ML) approach [5], [6]. Then, to guarantee well-conditioned solutions (in the Hadamard sense) to the TomoSAR nonlinear inverse problem, the derived ML-based approach is implemented in a closed fixed-point iterative adaptive manner, yielding the so-called MARIA (ML-inspired Adaptive Robust Iterative Approach) technique [5]. The use of statistical regularization approaches, within the maximum likelihood (ML) estimation theory, to solve the involved TomoSAR nonlinear ill-conditioned inverse problem, has been successfully demonstrated in the previous related studies [5], [6]. Within the main advantages of such approaches there is the retrieval of resolution-enhanced tomograms using a reduced (limited) number of passes, performing also suppression of artifacts and reduction of the ambiguity levels. Once the theoretical background of statistical regularization was provided, and its use for enhanced TomoSAR imaging was demonstrated, the subject of the work to be presented is focused on its application on different test sites and on the cross-check analysis of the retrieved measurements. References [1] A. Reigber and A. Moreira, “First demonstration of airborne SAR tomography using multibaseline L-band data”, IEEE Trans. Geosc. Remote Sens., vol. 38, no. 5, pp. 2142–2152, Sep. 2000. [2] F. Gini, F. Lombardini and M. Montanari, “Layover solution in multibaseline SAR interferometry”, IEEE Trans. Aerosp. Electron. Syst., vol. 38, no.4, pp. 1344-1356, Oct. 2002. [3] M. Nannini, R. Scheiber, and A. Moreira, “Estimation of the minimum number of tracks for SAR tomography”, IEEE Trans. Geosc. Remote Sens., vol. 47, no. 2, pp. 531-543, Jan. 2009. [4] M. Nannini, R. Scheiber, R. Horn, and A. Moreira, “First 3-D reconstructions of targets hidden beneath foliage by means of polarimetric SAR tomography”, IEEE Geoscience and Remote Sensing Letters, vol. 9, no.1, pp. 60-64, Jan. 2012. [5] G. D. Martín del Campo, M. Nannini, and A. Reigber, “Towards Feature Enhanced SAR Tomography: A Maximum-Likelihood Inspired Approach”, IEEE Geoscience and Remote Sensing Letters, pp. 1–5, August 2018. [6] G. Martín del Campo, A. Reigber and M. Nannini, “Feature Enhanced SAR Tomography Reconstruction through Adaptive Nonparametric Array Processing”, IEEE International Geoscience and Remote Sensing Symposium (IGARSS), 2018. [7] E. Aguilera, M. Nannini and A. Reigber, “A Data-Adaptive Compressed Sensing Approach to Polarimetric SAR Tomography of Forested Areas”, IEEE Geoscience and Remote Sensing Letters, vol. 10, no.3, pp. 543–547, Sept. 2012. [8] E. Aguilera, M. Nannini and A. Reigber, “Wavelet-Based Compressed Sensing for SAR Tomography of Forested Areas”, IEEE Trans. Geosc. Remote Sens., vol. 51, no.12, pp. 5283–5295, Dec. 2013.
Backscatter analysis using multi-temporal Sentinel 1 SAR data for irrigated crops monitoring in Banat region, Romania
Poenaru, Violeta; Badea, Alexandru
This study assesses the sensitivity of Sentinel C-band dual polarized SAR images acquired from March to September 2017 to monitor irrigated fields of maize, sorghum, soybeans and sunflowers in Banat, Romania. Physical process of the scattering mechanism occurring in crops in different phonological stages was investigated by assessing the temporal dynamics of the volume, double and odd bounce, entropy, anisotropy and alpha parameters. Additional information related to sub-canopy soil moisture was obtained by using coherent location of effective scattering centers within a resolution cell. After topography correction, SAR data were classified based on the distance of Wishart and information from coherence optimization was used. Availability of a dense time series of Sentinel 1 data improve monitoring of the soil parameters by considering temporal stability of the spatial soil moisture pattern. In this approximation soil roughness and vegetation conditions are not change and surface soil moisture can be estimated by using PolSAR and PolInSAR techniques. Preliminary results showed that the cross-polarized VH signal was useful for monitoring crops and soil conditions and was the least sensitive to differences in beam incidence angle.The VV polarized signal has a greater sensitivity to surface parameters and can be used for mapping of soil and crop cover characteristics.
Detecting aquaculture structures using the iDPolRAD and PINGPONG COSMO SkyMED
Marino, Armando; Ballester, David; Spyrakos, Evangelos
Aquaculture are a very valuable asset for many coastal countries and in the future they will play an important role in food security. Satellite remote sensing can improve the temporal and geo-spatial analysis of such marine facilities. Detecting platforms used for fish and shellfish farming provides a way to monitor assets and check they do not get damaged by storms. It also allows to identify illegal placement of structures in areas which should not host farms. In this work, we want to evaluate the potential of a new methodology that uses PolSAR data. This is called intensity Dual-Pol Ratio Anomaly Detector (iDPolRAD) and it was initially proposed to detect icebergs embedded in sea ice. Extensive work has been carried out on detecting ships using space-borne Synthetic Aperture Radar (SAR) systems. However, the identification of smaller and non-metallic targets is still challenging especially when the sea conditions are rough. This work presents the very first test of the iDPolRAD with aquaculture structures. The algorithms: The algorithm is based on the observation that the most of the maritime targets exhibit a different polarimetric behaviour compared to the sea. Specifically, the cross polarization channel and the ratio between cross- and co-polarizations (here referred to as depolarization ratio) increases. One of the reason is that complex targets (e.g. shellfish platforms) will provide scattering which will resemble Volume scattering or reflections from planes (mostly wet surfaces) with random orientations. They are therefore expected to have a polarimetric backscattering that is different from the one of the sea which is surface scattering. Two boxcar filters are applied over the HV and HH intensity images, exploiting two different window sizes: a smaller window or test window Wtest and a larger window or training window Wtrain. The detector can be writes as the different of the “HV filtered with the test window” minus “HV filtered with the train window”, all divided by the “HH filtered with the train window”. The previous operator is built as a ratio between intensities and therefore it is scale invariant. This is a very valuable property for a polarimetric observable, however scale invariance may be disadvantageous for some detection task. For instance, if the signal is very low and close to the noise floor, the detector may become noisy. An easy way out is by multiplying the detector by an intensity or magnitude image. In this context, the cross polarization channel should be preferred because it shows a higher contrast between icebergs and clutter. Multiplying the intensity of HV by the previous operator forms the iDPolRAD. If a pixel of the HV intensity image presents an anomaly in volume or oriented reflections, then it is multiplied by a large number. If it presents a homogeneous area, then it is multiplied by zero and if it presents a decrease in volume or oriented reflections, then it becomes negative. This enhances the contrast between anomalies in volume/oriented reflections and clutter. The data: In this work we tested the iDPolRAD on a dual-polarization HH/HV PINGPONG Cosmo-SkyMed dataset composed by 5 images of the coastal area near Vigo, Spain. This is an area intensely exploited for the production of mussels and hundreds of platforms cover the coastal area. Preliminary results: The iDPolRAD seems able to increase the contrast between the sea background and the platforms allowing the identification of more platforms. An extensive comparison of the dual-pol and single-pol algorithms will be presented. Reference: A. Marino, W Dierking, and C. Wesche, “A depolarization ratio anomaly detector to identify icebergs in sea ice using dual-polarization SAR images,” vol. 54, no. 9, pp. 5602–5615, 20
Polarimetric analysis of ALOS-2 SAR quad-pol data to detect icebergs
Bailey, Johnson Albert; Marino, Armando; Lanz, Peter
The presence of icebergs in cold waters represents a danger to navigation. Tracking and detection of icebergs has seen a lot of work carried out in the past decades using synthetic aperture radar (SAR) altimeters and scatterometers. However, previous work was limited in that the detection was dependant on iceberg size. Smaller icebergs continue to pose a challenge for detection, particularly those embedded in sea ice. The scattering behaviour of icebergs is the main focus of our work. Using 30 L-band ALOS 2 quad-pol images acquired from Antarctica, and the coasts of East and West Greenland, we conducted a polarimetric analysis using the application of several decompositions, including Cloud Pottier and those that are model based and coherent. Following the extensive analysis, the development of two new detectors was proposed to identify the specific scattering mechanisms that caused high levels of backscattering in icebergs. These maritime target detectors are the intensity Dual Pol Ratio Anomaly Detector or iDPolRAD (1) and the Polarimetric Notch Filter (2). The parameters of these two detectors are modified producing different versions that can be focused on specific scattering mechanisms, derived by the polarimetric analysis. Specifically the two images used in the iDPolRAD are selected considering the projection of the scattering vector over a) the dominant scattering mechanism representing the iceberg and b) an orthogonal vector. In terms of the Cloude Pottier decompositions these are the maximum and minimum eigenvectors over the iceberg area. The findings, as per our expectations, establish that icebergs have different polarimetric behaviours depending on conditions. However, the different polarimetric signatures seem to be well characterised by either volume or multiple reflections. The results of the detection exercise show that the detection performance is enhanced if we use the Cloude Pottier decomposition on historical data to extract the two most common iceberg signatures. This signature is then used to train two iDPolRAD algorithms and the final detection mask is the fusion of the detection masks obtained by subsequent runs of the iDPolRAD. Clearly, this is only possible when quad-pol data are available. We therefore show that performances can be improved through availability of quad-pol data. References: [1] A. Marino, W Dierking, and C. Wesche, “A depolarization ratio anomaly detector to identify icebergs in sea ice using dual-polarization SAR images,” vol. 54, no. 9, pp. 5602–5615, 2016. [2] A. Marino (2013) A notch filter for ship detection with polarimetric SAR data, IEEE Journal of Selected Topics in Applied Earth Observation and Remote Sensing, 6(3) pp. 1219–1232.
PyRAT: A Flexible SAR Postprocessing Toolbox
Reigber, Andreas; Martin del Campo Becerra, Gustavo; Jäger, Marc
PyRat (Python Radar Tools) is a flexible open-source framework for post-processing synthetic aperture radar (SAR) data. PyRAT is implementedin Python and distributed under an open-source license. It is mostly intended for post-processing of both airborne and spaceborne SAR imagery. PyRAT features an easy to use plugin-based programming interface, which allows users to quickly extend it with their own routines. PyRAT runs on various operating systems and is available free of charge on the web. The purpose of this paper is to give an overview of the current development status of PyRAT, in particular about the upcoming intergration with QGIS.
THE CO-POLARIZED PHASE DIFFERENCE OVER SEA SURFACE: THE INFLUENCE OF SAR ACQUISITION PARAMETERS AND ENVIRONMENTAL CONDITIONS
Buono, Andrea; de Macedo, Carina Regina; Nunziata, Ferdinando; Velotto, Domenico; Migliaccio, Maurizio
The exploitation of microwave remote sensing for marine and maritime applications is a hot topic in the scientific community. The almost all-weather capability of acquired large-scale information on the oceans independently of illumination conditions is a key benefit ensured by microwave remote sensing. Among the several microwave remote sensing tools, it was demonstrated that polarimetric synthetic aperture radar (SAR) can provide valuable information on the oceans once proper modeling is available. Actually, there is a wide set of marine and marine applications that takes full benefits of the polarimetric information, including ship detection, sea oil pollution monitoring, sea iceberg detection and coastline extraction [1-4]. Most of the polarimetric SAR-based algorithms that aim at detecting the target of interest needs to enhance the target-to-clutter ratio as much as possible in order to improve the detection accuracy. Nevertheless, the properties of sea surface polarimetric backscattering are strongly affected by both SAR imaging configuration (frequency, polarization, noise floor, incidence angle) and environmental conditions (sea state) [5-8]. In this study, one of the most widely and successfully used polarimetric features in the field of marine and maritime applications, namely the co-polarized phased difference, is considered [9-10]. A sensitivity analysis is undertaken on a large set of polarimetric SAR measurements collected, under different frequency, incidence angle and noise floor, over sea surface characterized by different sea state conditions, that aim at assessing the influence of SAR acquisition parameters and environmental conditions on the CPD distribution. The polarimetric SAR dataset includes high-quality wide swath full-polarimetric airborne UAVSAR scenes, full-polarimetric satellite Alos-PalSAR 1 scenes, full-polarimetric satellite Radarsat-2 scenes and dual-polarimetric coherent HH+VV satellite TerraSAR-X scenes. This analysis could be of potential support in the design and development of SAR architectures/algorithms whose goal is to identify sea targets of interest. [1] D. Velotto, F. Nunziata, M. Migliaccio and S. Lehner, “Dual-polarimetric TerraSAR-X SAR data for target at sea observation,” IEEE Geosci. Remote Sens. Lett., vol. 10, pp. 1114–1118, 2013. [2] D. Velotto, M. Migliaccio, F. Nunziata and S. Lehner, “Dual-polarized TerraSAR-X data for oil-spill observation,” IEEE Trans. Geosci. Remote Sens,, vol. 49, no. 12, pp. 4751–4762, 2011. [3] A. Marino, W. Dierking and C. Wesche, “A depolarization ratio anomaly detector to identify icebergs in sea ice using dual-polarization SAR images,” IEEE Trans. Geosci. Remote Sens., vol., 54, pp. 5602–5615, 2016. [4] F. Nunziata, A. Buono, M. Migliaccio and G. Benassai, “Dual-polarimetric C-and X-band SAR data for coastline extraction,” IEEE J. Sel. Topics Appl. Earth Observ. Remote Sens., vol. 9, pp. 4921–4928, 2016. [5] F. Nunziata, A. Buono and M. Migliaccio, “COSMO-SkyMed Synthetic Aperture Radar Data to Observe the Deepwater Horizon Oil Spill,” Sustainability, vol. 10, pp. 3599–3613, 2018. [7] A. Buono, F. Nunziata, C. R. de Macedo, D. Velotto and M. Migliaccio, “A sensitivity analysis of the standard deviation of the co-polarized phase difference for sea oil slick observation,” IEEE Trans. Geosci. Remote Sens., early access, 2018. [8] A. Buono, F. Nunziata and M. Migliaccio, “Analysis of Full and Compact Polarimetric SAR Features Over the Sea Surface,” IEEE Geosci. Remote Sens. Lett., vol. 13, pp. 1527–1531, 2016. [9] F. Nunziata, C. R. de Macedo, A. Buono, D. Velotto and M. Migliaccio, “On the analysis of a time series of X-band TerraSAR-X SAR imagery over oil seepages,” Int. J. Remote Sens, in print., 2018. [10] M. Migliaccio, F. Nunziata and A. Gambardella, “On the co-polarized phase difference for oil spill observation,” Int. J. Remote Sens., vol. 30, pp. 1587–1602, 2009.
Variable Reduction in Random Forests for Multi-sensor, Multi-temporal SAR Data Streams: Wetland Classification Case study
Behnamian, Amir; Banks, Sarah; White, Lori; Pouliot, Darren; Millard, Koreen; Pasher, Jon; Duffe, Jason - Environment and Climate Change Canada, Canada
Wetlands are important ecosystems for wildlife, as they provide habitat and services to many species; however, they are vulnerable to the cumulative effects of anthropogenic disturbance, pollution, and climate change. In light of this, many attempts have been made, especially over the past decade, to monitor dynamic changes operationally, via post-classification comparison with remote sensing data and, in particular, SAR data, which can be acquired regardless of cloud or haze. For this, classification algorithms such as Random Forests have been widely used with multi-sensor, multi-temporal data streams. The flow of frequent SAR data from different sensors with different polarizations, wavelengths, and incident angles allow for the possibility to generate hundreds of relevant predictor variables (e.g. C-band: quad polarimetry RADARSAT-2, dual polarimetry Sentinel-1, and compact polarimetry RCM; X-band: TerraSAR-X; and L-band: ALOS-2). These can range from backscatter intensity channels, to decompositions of backscatter matrices, and/or derivatives of digital elevation products, all of which can provide useful and complementary information. With Random Forests, users also have the advantage of ranking predictor variables by their relative importance using measures such as Mean Decrease in Accuracy or Mean Decrease in Gini index. However, in the presence of multiple correlated variables, which is often the case for multi-sensor, multi-temporal data sets, rankings become unreliable. This is because of the so called spreading of importance among correlated variables, which results in a bias toward those provided without surrogates. A possible solution is to identify and remove correlated variables; however, this task is complex for SAR data streams since, in many cases, the association between variables cannot be explained with a simple linear correlation. As a result, the use of measures such as correlation coefficient is ineffective for identifying those variables providing redundant and therefore interchangeable information. In this study, we evaluate whether a simple statistical measure of association, which can be interpreted as a (statistically) low order measure of class separability, is more effective for reducing the number of predictor variables. The method, referred to as the point biserial correlation, shows the strength and direction of association between the training dataset (i.e. categorical data) and predictor variables (i.e. continuous data). The measure of association is evaluated for all possible combinations of class pairs, for all training points, for each predictor variable. The method was applied for classifying wetlands over a study site in the Bay of Quinte, Ontario, Canada, using two datasets. The first dataset consists of different combinations of multi-temporal and multi-angle quad pol RADARSAT-2, simulated compact pol RADARSAT Constellation Mission (RCM), and fine and coarse resolution DEM and DSM information used to classify shallow water, marsh, and swamp and as well as forest, and agriculture/non-forest. The second dataset consists of a combination of ALOS-2, RADARSAT-2, TerraSAR-X, and DEM over the same study area with similar classes. Landsat-8 optical data was also added to the latter dataset to investigate its effects in comparison to only multi-sensor SAR data. Results show that from 198 generated predictor variables from the latter, our method can shortlist four variables that offer approximately the same per class accuracies that can be achieved with at least 15 variables if users use a Mean Decrease Accuracy measure (of Random Forests) to reduce the number of variables. The effects of such a difference between the minimum numbers of required variables for maximizing per class accuracies are discussed in terms of computational expense for national scale wetland classification, and suggestions are made on how to use the algorithm in an operational setting.
Analysis and Interpretation of Polarimetric L-band Ground-Based SAR Echoes from a Partly-Vegetated Land Surface
Özdemir, Caner; Demirci, Sevket; Isıker, Hakan; Yilmaz, Betul; Gokkan, Serhat - Mersin University, Turkey
Microwave backscattering is highly sensitive to target as well as radar parameters which prevents, in many instances, the derivation of definite conclusions. Ground-based synthetic aperture radars (GB-SAR) are utilized as a useful and practical tool in gaining an improved understanding of this complex mechanism. They also provide complementary information on evaluating the validity of the polarimetric analysis results of air or satellite-based SAR applications. For this reason, various GB scatterometers, imaging radars and interferometers have been studied and evaluated. In this study, we present the results of our investigation of the polarimetric properties of L-band GB-SAR backscattering from a typical partly vegetated land surface as well as some man-made targets. A terrain surrounding a building was chosen as a test field and the radar system was located on its roof terrace. The target area of interest was comprised of a vegetated land, dirt roads, a parking lot, various man-made items and add-on trihedral corner reflector (TCR) targets. The TCRs were placed onto different ground sections and inclined with a zero angle w.r.t. the bottom surface. Among the man-made objects were a car, a lighting pole, a signboard, two tall vertical metal objects and various small sized materials located mostly within the parking lot. The zone covered by vegetation was a scattered compound of grasses, plants, trees and bare soil areas. The SAR system was mounted on a wheel-based moving platform and composed of a vector network analyzer, two linearly polarized Vivaldi horn antennas and a 1-watt RF amplifier. The antennas were arranged in a quasi-monostatic configuration with 30 cm spacing and located at a height of h=24 m with 9° inclination angle from the horizontal direction. The synthetic aperture length was set to 13.1 m and sampled with ∆x=10 cm steps. The backscattered data were acquired for full linear polarimetric modes (HH/HV/VHVV) and for a frequency span of 0.75 GHz to 2.25 GHz with 1601 sampling points. The collected data were then focused with a near-field back-projection imaging algorithm to account for the spherical wave-fronts. The polarimetric GB-SAR image signatures are analyzed and interpreted via examination of; (i) intensity images in terms of the backscattering mechanisms and their dependency on frequency, polarization and look-angle, (ii) total power image of the channels, (iii) Pauli RGB images, (iv) distribution of values in H/alpha space for specific target regions, (v) H/alpha classification and (vi) beta-orientation angle images. The intensity images show successful detection and imaging of the terrain features such as dirt roads, trees, vegetation as well as the man-made objects like lighting pole, cars, corner reflectors etc. While man-made objects can be seen in both polarizations, natural terrain features such as leaves, trees and bush are more pronounced in cross-polarized GB-SAR images. The polarimetric analysis results are also found to be in good agreement with the predicted scattering mechanisms from natural and manmade objects. Specifically, H/alpha classification results are shown to be capable of clearly identifying the distinct scattering mechanisms present in the scene.
3-D study of a Random Volume over Ground: underlying ground characterization using controlled PolTomSAR experiments
Abdo, Ray; Ferro-Famil, Laurent
I. INTRODUCTION This paper aims to evaluate the potential of Polarimetric SAR Tomography (PolTomSAR) [1] for analyzing a semi-opaque Random Volume lying over a rough Ground (RVoG), and to assess different characterization methods as well as ground and volume separation methods, such as the Sum of Kronecker Products (SKP) and Hybrid SKP (HySKP), or Full-Rank 3-D polarimetric imaging [2][3]. This study is based on the use of controlled 3-D experiments, during which a simplified and miniaturized RvoG-like scene is imaged with a laboratory 3-D SAR, operated along a 2-D aperture. Polarimetric and tomographic signals are acquired for various configurations, and different scattering mechanisms are naturally captured by removing, hiding or replacing different parts of the scene. This unique possibility to isolate specific scattering mechanisms, compared to acquisitions performed over natural environments, through PolTomSAR is used to evaluate the performance and relevance of existing decomposition techniques generally applied to forest characterization. The final objective of this series of measurements concerns the characterization of the underlying ground, in terms of roughness, humidity... II. EXPERIMENT DESCRIPTION The Ground-Based SAR system used in this study has been developed at the IETR, University of Rennes 1, France. It is based on the use of a Vector Network Analyzer, and is placed on a moving platform with an effective azimuthal aperture of 3m. In the frame of this study, it is operated at X-band (fc= 10GHz), over a bandwidth of 4GHz and around an incidence angle of 45◦, providing SAR images with an azimuth resolution of δaz = 1.73cm and a range resolution of δrg = 3.75cm. Using an irregular array of 4 antennas, together with bi-directional acquisition capabilities, the GB PolTomSAR generated a virtual regularly spaced array of 6 antennas operating in single pass mode [4]. Additional baselines may be created by shifting the system in the vertical direction over several passes. The first measured scene contains sand on smooth ground to simulate a low roughness and low humidity ground, with a dielectric constant of ε=4. The knowledge of the true ground response is a major advantage of a controlled scene. The second scene called volume over ground contains a volume with anisotropic particles shifted in elevation using a polyester pole. The altitude of the volume is large enough so that side-lobes of its tomographic response do not interfere with the ground scattering. In this scene, three main mechanisms are identified, single-bounce on ground, single-bounce on volume and double-bounce ground/volume which is focused at the ground height. As a consequence, by comparing the scene’s ground response, isolated with 3-D focusing, with the reference ground scene, and since the volume’s altitude is sufficiently high to avoid interferences with the ground, we are able to quantitatively characterize the double-bounce mechanism, and validate or infer scattering hypotheses, such as the ones discussed in [5] and [6]. In the third scene, wood stems and dipoles replace the polyester pedestal. This scene called volume over trunk over ground shows the difference between ground/volume and ground/trunk double-bounce scattering contributions. It follows that, in order to estimate the bare soil contribution gs, we need to subtract the double-bounce contributions gd from total ground response, i.e. gs = gt − gd. This is done using the SKP or HySKP methods introduced respectively in [7] and [8]. After that, a comparison between ground contributions found in these experiments with the ground only scene estimate the reliability of these methods, and an alternative technique is proposed. REFERENCES [1] A. Reigber and A. Moreira. First demonstration of airborne sar tomography using multibaseline l-band data. IEEE Transactions on Geoscience and Remote Sensing, 38(5):2142–2152, Sept 2000. [2] Laurent Ferro-Famil, Bassam El Hajj Chehade, Ray Abdo, Dinh Ho Tong Minh, Stefano Tebaldini, Thuy Le Toan ; IMPROVED CHARACTERIZATION OF A TROPICAL FOREST USING POLARIMETRIC TOMOGRAPHIC SAR DATA ACQUIRED AT P BAND ; [International Geoscience and Remote Sensing symposium- IGARSS] (2018) ; [3] L. Ferro-Famil, Y. Huang and S. Tebaldini, "Polarimetric characterization of 3-D scenes using high-resolution and Full-Rank Polarimetric tomographic SAR focusing," 2016 IEEE International Geoscience and Remote Sensing Symposium (IGARSS), Beijing, 2016, pp. 5694-5697. doi: 10.1109/IGARSS.2016.7730487 [4] Temesgen Gebrie Yitayew, Laurent Ferro-Famil, Torbjørn Eltoft, and Stefano Tebaldini. Tomographic imaging of fjord ice using a very high resolution ground-based sar system. IEEE Transactions on Geoscience and Remote Sensing, 55(2):698–714, 2017. [5] N. Lahlou, L. Ferro-Famil and S. Allain-Bailhache, "Retrieving soil moisture below a vegetation layer using polarimetric tomographic SAR data," 2014 IEEE Geoscience and Remote Sensing Symposium, Quebec City, QC, 2014, pp. 3239-3242. doi: 10.1109/IGARSS.2014.6947169 [6] N. Lahlou, L. Ferro-Famil and S. Allain-Bailhache, "Study of soil respons under a vegetation layer using TomSAR data and ground-based TomSAR data," 2015 IEEE International Geoscience and Remote Sensing Symposium (IGARSS), Milan, 2015, pp. 1324-1327. doi: 10.1109/IGARSS.2015.7326019 [7] S. Tebaldini, "Algebraic Synthesis of Forest Scenarios From Multibaseline PolInSAR Data," in IEEE Transactions on Geoscience and Remote Sensing, vol. 47, no. 12, pp. 4132-4142, Dec. 2009. doi: 10.1109/TGRS.2009.2023785 [8] M. Pardini and K. Papathanassiou, "On the Estimation of Ground and Volume Polarimetric Covariances in Forest Scenarios With SAR Tomography," in IEEE Geoscience and Remote Sensing Letters, vol. 14, no. 10, pp. 1860-1864, Oct. 2017. doi: 10.1109/LGRS.2017.2738672
An investigation of the effect of polarimetry on forest structure estimation using TanDEM-X high resolution digital elevation models
Choi, Changhyun; Pardini, Matteo; Papathanassiou, Konstantinos
TanDEM-X forms a spaceborne X-band single-pass interferometer together with TerraSAR-X, allowing the acquisition and analysis of interferometric synthetic aperture radar (InSAR) data in forested area globally without temporal decorrelation. For this reason, the exploration of interferometric TanDEM-X data for forestry has been increasing in time especially concerning forest height estimation [1-3], biomass estimation and classification [3-5]. More recent experiments have demonstrated that TanDEM-X data can contribute to the quantification of forest structure [6-8], although this potential is today only partially characterized. Forest structure is linked to the three- dimensional (3-D) distribution of vegetation elements in a forest. It is commonly accepted that forest structure can be quantified by mean of two indices expressing heterogeneity in the horizontal and vertical dimensions. While the horizontal structure index reflects stand density of trees, the vertical one accounts for tree size variability. Recently, indices aiming at describing the same structural information have been developed using SAR reflectivity profiles estimated from sets of SAR images acquired under slightly different incidence angles by means of tomographic techniques [9]. However, such framework cannot be applied to TanDEM-X data, as suitable tomographic datasets are available only in a very limited number of test sites. Thus, in the most general case, forest structure information has to be extracted from one interferometric acquisition only. One possibility is to use the InSAR coherence phase to calculate a high resolution digital elevation model (DEM). Then, a height profile can be derived by computing the histogram of the DEM heights within a unit scale and such profile can in turn be used to extract structure information. Due to the limited X-band penetration capabilities, these profiles still reflect horizontal heterogeneity as, for instance, they allow extracting information about local variabilities of top canopy height [8]. The purpose of this work is to investigate the role of polarimetry on forest horizontal structure estimation from TanDEM-X DEM height profiles. Indeed, the reflectivity profiles underlying InSAR coherence may vary with the polarizations affecting the related DEM height. Height profiles calculated from different polarimetric channels and the related horizontal structure information are compared. In particular, it is investigated if the polarization diversity at X-band can be exploited for coherence optimization for the extraction of more accurate structure information. It is expected that the results of these comparisons will depend on X-band penetration capabilities and its characteristic related to the specific site [1]. In order to assess this, experiments will be carried out by processing dual-pol (HH/VV) TanDEM-X data acquired over different tropical and temperate forest sites. Comparisons with structure metrics calculated from Lidar and field inventory data will be presented as well. Reference (1) Kugler, Florian, et al. "TanDEM-X Pol-InSAR performance for forest height estimation." IEEE Transactions on Geoscience and Remote Sensing 52.10 (2014): 6404-6422. (2) Qi, Wenlu, and Ralph O. Dubayah. "Combining Tandem-X InSAR and simulated GEDI lidar observations for forest structure mapping." Remote sensing of Environment 187 (2016): 253-266. (3) Persson, Henrik J., et al. "Experiences from large-scale forest mapping of Sweden using TanDEM-X data." Remote Sensing 9.12 (2017): 1253. (4) Solberg, Svein, et al. "Estimating spruce and pine biomass with interferometric X-band SAR." Remote Sensing of Environment 114.10 (2010): 2353-2360. (5) Caicoya, Astor Toraño, et al. "Large-scale biomass classification in boreal forests with TanDEM-X data." IEEE Transactions on Geoscience and Remote Sensing 54.10 (2016): 5935-5951. (6) De Grandi, G. D., et al. "Analysis by Wavelet Frames of Spatial Statistics in PALSAR Data for Characterizing Structural Properties of Forests." (2009). (7) Pulella, Andrea, et al. "Tropical forest structure observation with TanDEM-X data." 2017 IEEE International Geoscience and Remote Sensing Symposium (IGARSS). IEEE, 2017. (8) Choi, Changhyun, et al. "GEDI – TanDEM-X fusion for enhanced forest structure observation: A comparison of InSAR height profiles and lidar full waveforms." FORESTSAT 2018; 8th Association for Forest Spatial Analysis Technologies, 2018. (9) Tello, Marivi, et al. "Forest structure characterization from SAR tomography at L-Band." IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing 99 (2018): 1-13.
Decision Tree Classification Method based on Adaptive Dimensional Polarimetric Features of PolSAR Data
Cheng, Jianda (1); Yin, Qiang (1); Zhang, Fan (1); Hong, Wen (2) - 1: Beijing University of Chemical Technology, China, People's Republic of; 2: Institute of Electronics, Chinese Academy of Sciences
Decision tree method is applied to PolSAR data classification, due to its capability to interpret the scattering characteristics as well as good classification accuracy. At present, the decision tree based classifiers widely applied to PolSAR image classification often adopt one single polarization features at the nodes of the decision tree. Some studies employ the two-dimensional polarization features. As the dimension of features increase, the accuracy of classification is improved. In this study, the dimension of features used in the decision tree nodes reaches three, and fisher linear discriminant is used to reduce the dimension of decision tree nodes. Increasing the dimension of decision tree nodes inevitably leads to computational burden. In fact, not all nodes need three-dimensional features to achieve high-precision classification. Therefore, an adaptive dimension of the decision tree is introduced in this paper. The dimension of the nodes starts from one dimension and expands to two or even three dimensions if needed. On one hand, the three-dimensional features are used to improve the classification accuracy, and on the other hand, the computational complexity is reduced by the adaptive dimension. At the same time, the scattering mechanism of categories is described more specifically. The proposed classification scheme is verified by the classic PolSAR data in flevoland area, as well as GF-3 data. It is superior to the traditional one-dimensional and two-dimensional decision tree methods.
Hybrid-pol data analysis for land cover classification
Kumar, Ajeet
Kumar, Ajeet (1); Sharma, Rakesh (1); Das, Anup (2); Panigrahi, Rajib Kumar (1) - 1: IIT Roorkee, India; 2: Space Application Center (ISRO)
For many years, the full polarimetric (also known as full-pol or quad-pol) Synthetic Aperture Radar (SAR) systems have provided promising results for Earth observation purposes. However, the full-pol system suffers with poor revisit time. As a result of which, it is not preferable for the applications requiring frequent Earth observation. The revisit cycle can be made shorter by increasing the swath coverage area of the satellite. The dual-pol SAR systems have the advantage of half pulse repetition frequency (PRF) than the full-pol systems, that leads to increased swath coverage and decreased data rate. A coherent dual-pol SAR system, popularly known as compact-polarimetry (compact-pol) SAR system, retains the relative phase difference between the two received channels to fully characterize the backscattered field. In recent times, it has been shown that the results obtained using compact-pol SAR is almost comparable with full-pol for certain scenes and applications. There are various compact-pol systems based on different transmitted polarization and coherent dual linear receive polarization. Among all compact-pol modes, hybrid-polarimetry (hybrid-pol) has the optimum architecture as established in literature. Hybrid-pol SAR system transmits circular polarization and coherently receives the two linear orthogonal polarizations. Radar imaging satellite-1 (RISAT-1) was the first satellite that provided hybrid polarimetric (hybrid-pol) data for the Earth observation purposes. The C-band Radar Imaging Satellite-1 (RISAT-1) was launched by Indian Space Research Organization (ISRO) in April 2012. The processing of RISAT-1 data can be carried out using two different types of approaches. First, by reconstructing 3$\times$3 pseudo quad-pol information from 2$\times$2 hybrid-pol measurement. After reconstruction, any well established full-pol decomposition techniques can be implemented. Second, by using decomposition techniques which directly take 2$\times$2 hybrid-pol measurement as an input. The first type of approaches facilitate us to utilize the rich literature of full-pol systems. However as established in literature, these approaches may not be suitable for land-cover classification in the area that has elevation variations. Contrarily, second type of approaches can work well in the areas with slopes or having elevation-variations which can lead to better land-cover classification results. The both types of approaches are frequently reported in literature for the analysis of hybrid-pol data synthesized from full-pol data. Using synthesis process, the hybrid-pol data of pure-circular transmission cannot be generated. This may affect the accuracy of final-classification results. Hence, to explore the actual potential of hybrid-pol system, it is needed to use original hybrid-pol data. In this paper, we investigate the potential of hybrid-pol system for land-cover classification using actual hybrid-pol data of RISAT-1 satellite acquired over Mumbai (India) city. The analysis of RISAT-1 data is carried out using different types of decomposition techniques where each land-cover scatterers are classified in any of the three basic scattering mechanisms viz. single-bounce indicating Bragg surface region, double-bounce indicating urban region and volume scattering indicating forest/mangrove region. By comparing the performance of decomposition techniques on the basis of land-cover classification accuracy, one can discover that which (First or Second) type of hybrid-pol analytical techniques are more suitable for the processing of RISAT-1 datasets.
InSAR coherence-based land cover classification of Okara, Pakistan
Khalil, Rao Zahid; Haque, Saad Ul - Institute of Space Technology, Pakistan
Reliable and current availability of land cover knowledge are essential for many studies regarding planning, management, monitoring and updating activities. The optical satellite sensor data has been utilized for the classification of land use/land cover. In this study, the capability of synthetic aperture radar (SAR) interferometric coherence is practiced for land cover classification in Okara, Pakistan using Sentinel-1A imagery. Two Single Look Complex (SLC) product of months April and May 2016 were used and processed to create backscatter and interferometric coherence layers. From backscatter layers of each month, the mean backscatter and backscatter difference layer were obtained. False color composite (FCC) were developed comprising mean backscatter, backscatter difference and coherence, and performed supervised classification using maximum likelihood method to generate land cover classes i.e. water, barren, vegetation and built-up. Kappa statistics were employed for accuracy assessment of the output map. Results showed the good potential of Sentinel-1C-band for land cover classification having 0.69 Kappa coefficient and 80% overall accuracy. This study investigated the potential of C-band backscatter coefficients and coherence map for land cover discrimination. Coherence proved to be efficient in the examination of vegetative and non-vegetative areas.
SAR images compressed sensing based on Orthogonal Matching Pursuit algorithm
Roubah, Slim; Ouarzeddine, Mounira; Souissi, Boularbah
Synthetic Aperture Radar (SAR) is an active remote sensing system capable of producing high-resolution images. The huge amount of generated samples implies high energy costs for processing, storage and transmission. Compressed Sensing (CS) theory suggest that a signal can be recovered from a few measurements below the Nyquist Rate. The basis where the processed signal can be sparse must be known in order to apply CS. The standard transformations like FFT, DCT and wavelets may give a bad reconstruction especially in the SAR domain. To avoid a basis heavy computing for every signal, an alternative way to find the sparsity is proposed in this paper. The recovery algorithm is used in the compression step to compute a signal sparse representation using the Fourier transform. It can take more processing time, but it ensures a sparse vector representation and a good reconstruction. This method is tested on Flevoland SAR image, sparse representations and recovered images are presented to show the proposed process performance and evaluated using the MSE, PSNR and correlation parameters. An exploitable image can be reconstructed from 50% of data.
Classification of Forest Areas with Multi-Frequency Fully Polarimetric and Multi-Baseline Interferometric SAR Data: the Haspelmoor Case Study
Persico, Federica; Soprano, Isabella; Joerg, Hannah; Pardini, Matteo; Hajnsek, Irena; Papathanassiou, Kostas; Iodice, Antonio
Synthetic Aperture Radar (SAR) Polarimetry and Interferometry can be used to characterize the electromagnetic scattering from natural targets and to compute important geophysical parameters. These techniques are exploited here to investigate scattering differences over a flooded area at different frequencies and with different baselines.The analysis was carried out using fully polarimetric and multi-baseline SAR data acquired by DLR’s airborne system F-SAR over the Haspelmoor testsite in the South-East of Germany in June 2016. The testsite is characterized by forest and fields either with flooded ground (“Moor”) or with dryer ground. The entropy-alpha plane [1] can be used for extraction of physical information about the scattering mechanisms and it allows to perform a very simple classification of the data. However, depening on the frequency, the physical meaningfulness of these initial classes might not be sufficient and targets with different vertical structure properties could be mixed in one class. The main objective of this paper is to understand the added value of using (1) multiple frequencies or (2) interferometric coherences in order to classify forest and wetland areas. In the first part of this work the iterative Wishart classification is applied at different frequencies using the classes from the entropy-alpha plane as an initialization [2]. Using two different initializations (the entropy-alpha and a random noise) it was found that the initialization does not influence the final classification results. To get a better idea of how the classes at different frequencies are related between each other, a confusion matrix is used. In particular, C- and S-band have almost the same number of pixels in the same class, while X- and L-band show much bigger differences. For instance, there are classes at L-band which can be divided in two or more classes at X-band, especially over the forest. In the second part, it is investigated if the classification benefits from additional information about vertical structure, i.e. from using interferometric coherences. The iterative Wishart Classifier is applied to interferometric coherences starting from a polarimetric initialization (entropy/alpha segmentation) [2] at L-band where the polarimetric classification was not very distinguished, particulary over forest. Analyzing the results as a function of the vertical wavenumber shows that classes containing mainly “Forest” and “Fields” preserve their meaning as the wavenumber varies, but for large values the classes’ separability reduces and classification errors increase. Further, forest classes seem to be more separable than field classes. In order to understand the number of classes which is physically sufficient, i.e. which classes could be merged, the possible objects in the scene have been divided in three macro-classes (Forest, Field, Else) and the statistical separability of the different classes has been evaluated [3]. The classes which have been judged not separable, yielded two classes within each macro-class. In the last step the classification on this reduced number of classes has been performed again. In each macro-class the separability with respect to the sub-class extension has been computed in order to understand if the classes make sense physically. Finally, comparing the polarimetric and interferometric approaches in the L-band case, the results show that the Forest classes are more separable and the interferometric information allows to obtain a better (more detailed) classification in the vegetated areas. Moreover, there is a correspondence between the classes which represent surface scattering. [1] S.R.Cloude and E.Pottier. An entropy based classification scheme for land applications of polarimetric SAR, IEEE Transactions on Geoscience and Remote Sensing, vol.35, n°1,pp. 68-78, January 1997. [2] J.S. Lee, M.R. Grunes, T.L. Ainsworth, L.J., Du, D.L. Schuler, and S.R. Cloude. Unsupervised classification using polarimetric decomposition and the complex Wishart classifier, IEEE Transactions on Geoscience and Remote Sensing, vol 37/1, n°5, p 2249-2259, September 1999. [3] J.S. Lee and E.Pottier. Polarimetric Radar Imaging: From Basics to Applications. Optical Science and Engineering. CRC Press, 2009.
The significance of transmitting a circular polarized component over a linear polarized component of RISAT-1 data for land cover classification
DASARI, KIRAN; LOKAM, ANJANEYULU - NATIONAL INSTITUTE OF TECHNOLOGY WARANGAL, INDIA
In this paper, we have accessed the performance of circular transmitting polarization (hybrid polarimetry) SAR over linear transmitting SAR (dual polarimetry) for land cover classification. The uniqueness in transmitting a circular polarization helps us to achieve larger swath width with low power consumption, compared to quad polarimetry SAR systems. This paper provides a comparison of the information content of hybrid polarization over dual polarization of RISAT-1 data. Few studies have compared the simulated hybrid pol data with dual pol data. In this paper, the hybrid pol data and dual pol data are acquired from the same mission (RISAT-1) with same resolution. In this study, we have identified five land cover classes (Urban, water body, paddy, cotton, Mango plantation) based on the ground truth data. Visually, we can compare the difference between circular transmitted component and linearly transmitted component on raney decomposed RGB image. Hybrid pol data has discriminated various land cover very well compared to dual pol data. Supervised classification was performed using SVM classifier with (RBF) kernel. We have obtained an overall accuracy of 80.62% for hybrid pol dataset and 54.3% for dual pol dataset. The classified results were validated using ground truth data, optical sensors data (Resourcesat-2) and Google Earth.
Incoherent Polarimetric target scattering decomposition: A overview and their implementation in TerraLib system
Carvalho, Naiallen; Sant'Anna, Sidnei; Bins, Leonardo
The Target Decomposition of Fully Polarimetric Synthetic Aperture Radar (PolSAR) Images, which consists in identify and isolate different scattering mechanisms based on the scattering matrix decomposition, is an important field of study in the targets characterization and feature extraction. Generally speaking, the target decomposition could be split into two main groups: coherent decomposition and incoherent decomposition. The so-called coherent targets are those which doesn’t depolarize the wave when interacting with it, such as corner reflectors used for radar calibration. For this kind of situation, the target decomposition decomposes the scattering matrix into a sum of complex elements and each element represents a certain canonical scattering mechanism. Examples of coherent target decomposition are Cameron decomposition and Krogager decomposition. On the other hand, the incoherent targets are those with random backscatter, which produces partially polarized waves or completely depolarized waves, as natural targets in the scene (trees or buildings). These kinds of the target are analyzed by the statistical point of view, therefore they are derived by the covariance matrix, or by coherence matrix, or by the equivalent 4 x 4 Müller matrix. The incoherent target decomposition can be split into two groups: the first one are the decompositions based in models, such as the Freeman-Durden decomposition and Yamaguchi decomposition, and the second group is the ones based on eigenvalues and eigenvectors analyzes, such as Cloude-Pottier decomposition and Touzi decomposition. In this work we choose to compare the three-component scattering model decomposition by Freeman and Durden due to its effectiveness in providing scattering powers for each scattering component in natural distributed areas, and the three-component method based on an eigenvalue analysis proposed by Cloude and Pottier, which is a method for extracting the average parameters from data using a smoothing algorithm based on second-order statistics with the goal of to find the dominant scattering mechanism via extraction of the largest eigenvalue. These two algorithms were implemented using C++ language on the framework TerraRadar, which is based on TerraLib Library, a GIS (Geographic Information System) component library developed by the National Institute for Space Research (INPE) and contributors. The tested images were obtained by PALSAR satellite, being a cut over part of the Tapajós National Forest, an important conservation unit in the Brazilian Amazon. We present a short analysis of both theorems and its importance regarding the backscatter answer from terrain targets.
A wetland mapping method by temporal integrals derived from H/A/alpha decomposition of Sentinel-1 images
Pacskó, Vivien; Petrik, Ottó; Friedl, Zoltán; Nádor, Gizella; Kristóf, Dániel; Belényesi, Márta; Molnár, Gábor
Wetlands are dynamic and diverse ecosystems, and play important role for example in decreasing the likeliness of floods or in filtering, and cleaning of surface and waste water. The identification and monitoring of them is essential to preserve them as ecosystem service providers. For reaching good classification result with an ensemble learning method like Random Forest, the selection of input data is fundamental. When the land cover is as complex as wetlands, and the collected reference data is uncertain or not up-to-date, selecting the best remote sensing input layers becomes even more significant. So the intent of my study is to find proper temporal integrals of polarimetric descriptors that could support a supervised classification in finding wetlands. The studied time period is from October, 2014 to October, 2018, that means more than a hundred acquisition dates. Polarimetric descriptors were generated based on H/A/Alpha decomposition of covariance matrix of Sentinel-1A dual-pol (VV+VH) data. From the downloaded SLC images we extracted scattering coefficients (Sigma0), and derived the following ten descriptors: alpha and its two components, anisotropy, entropy, the two eigenvalues of covariance matrix, Shannon entropy and its two components. In this study, the examined area is one from the Kiskunság region, Hungary, having different types of wetlands, grasslands, scrubs, and agricultural habitats. The reference data is provided by Ministry of Agriculture, Department for Nature Conservation, and its categories correspond to the General National Habitat Classification System.
Improved wetland vegetation mapping by using bistatic coherence of the TanDEM-X mission.
Mleczko, Magdalena; Mróz, Marek; Fitrzyk, Magdalena
As a continuation of previously published research paper concerning the use of Sentinel-1 and TerraSAR-X/ TanDEM-X (TSX/TDX) data for wetland mapping (Remote Sens. 2018, 10, 78; doi:10.3390/rs10010078) the authors propose to investigate further the input of bistatic coherence data into the increase of classification accuracy and wetland mapping reliability. For the research led on the Biebrza wetlands in N-E Poland multiple TSX/TDX images have been acquired during the Science Phase (2014-2015) both in monostatic pursuit or bistatic formation of the TSX/TDX constellation. The vegetation types and its succession resulting from long- or short-term flooding on the wetland were the object of investigations and mapping. The open water surface was the object of mapping as well. The multitemporal approach based on exploiting: i) amplitude data captured at different polarizations and at different incidence angles, ii) decomposition of quad-pol datasets, iii) Shannon Entropy images for mapping purposes have been adopted. The published results proved that TSX/TDX time-series permitted to extract main vegetation types of this herbaceous wetland and partially submerged vegetation as well. Multitemporal coherence images turned out to be useful in surface water extent mapping but not in vegetation study. In the presented paper we focus on bistatic coherence data as an additional feature complementing amplitude(s) in “classification space”. We noticed that bistatic coherence has been highly preserved for some vegetation types (e.g. large sedge surfaces) despite of expected volume decorrelation effects. These effects were normally observed for the forest areas and surprisingly for built-up areas as well. Object-based and support vector machine (SVM) classifiers were used for classification purposes. The validation of the results was made based on optical images and UAV reconnaissance over the area of interest.
Monitoring agricultural fields in South America using Sentinel 1: The Asparagus case study in Peru
Silva, Cristian; Marino, Armando; Cameron, Iain
According to the Food and Agriculture Organization FAO of the United Nations, Peru is the world’s second biggest asparagus producer. It is the third most exported agricultural product of Peru after grapes and mango [1,2]. The present study evaluates the viability of monitoring extensive asparagus fields in Peru using the capabilities of the Sentinel 1 satellite. Ground data acquired during two campaigns a year for 2017 and 2018 has been used to analyse the radar response with respect to the different crop growth stages. The measurements on the field provide a tool to make an interpretation of the single look complex images acquired and processed over the study area. Several parameters of the radar satellite response have been analysed, including the VV and VH backscatter as well as its ratio. Similarly, the parameters obtained after performing the dual polarimetric decompositions such as the alpha/entropy decomposition are analysed against the ground truth provided by the farmers. Preliminary results confirm the potential of sentinel 1 for monitoring the fields. The temporal trends of the asparagus crop radar backscatter indices can be correlated to the phenological stages provided by the ground truth. The following statements have been tested and confirmed: • The HV channel shows remarkable sensitivity to the vegetation growth. • The backscatter ratio shows correlation with the presence of vegetation in the fields • The alpha angle obtained after dual polarization decomposition allows to clearly determine moments of harvest and the beginning of a new season. • Deviations from the normal behaviour of the time series can be visually identified, therefore, they can be automatically detected. This was proved, for instance, with abnormal peaks in the VV channel during a torrential rain period in the study area. • The temporal trends of the different variables present a repetitive behaviour during the four seasons considered, thus, enabling the possibility to create crop models that replicate the crop evolution based on historical satellite and ground truth data. • Simplified manual algorithms show the potential to retrieve the most important crop stages, such as the beginning of a new season, the vertical growth of asparagus spears, the establishment of a fern and the harvest dates. Using these preliminary findings, current research focuses on the design of an automatic classification tool to retrieve the crop growth stage, based on the ground and satellite data from previous campaigns. Additional polarimetric features are also being tested and will be included in the tool in order to increase the number of crop stages retrieved as well as the accuracy of its performance. Additionally, an adaptation of a quad polarimetric change detector has been done to the case of dual pol data, with the aim of evaluating the radar changes caused by the changes on the structure and dielectric constant of the crop during its evolution [3]. Multitemporal analysis based on change detection facilitates the understanding of the radar backscatter response [4]. Acknowledgement This research was supported by the Project EO4cultivar, led by Environment Systems Ltd and co-funded by the UK Space Agency. The EO4cultivar project is an international collaboration project with partners in the UK, Peru and Colombia through the International Partnership Programme (IPP) of the UK Space Agency.
Review of Classification Methods and Algorithms for Optical and Radar Remote Sensing Images
El kharki, Omar; Mechbouh, Jamila - Faculty of Sciences and Techniques of Tangier, Tangier, Morocco, Morocco
The classification of optical or radar remote sensing images is the process of grouping the pixels of this image into a limited number of classes. Classification of remotely sensed data has attracted the attention of the remote-sensing community because classification results are the basis for many environmental and socioeconomic applications. In recent years, in optical remote sensing, many advanced classification approaches, such as artificial neural networks, decision trees, fuzzy sets, have been widely applied for image classification. Each classification method has its own merits. Selecting an appropriate classification approach for a specific study is not easy. Different classification results can be obtained according to the selected classifier(s). in the first part of this paper, we review various methods of classification of optical remotely sensed data with an analysis and comparative study. In radar remote sensing, the wave is polarized and affected by an inherent noise called speckle, thus classical remote sensing algorithms can’t be directly used. In the second part of this paper, we present a panoramic view about polarimetric decompositions in order to extract the principal mechanisms of backscattering (Odd‐bounce, Even‐bounce, Oriented Even‐bounces, etc) from radar images. Finally, the main classification methods adopted for this type of imagery are presented.
Assessment of a Ku-Band polarimetric bistatic terrestrial radar: KAPRI
Stefko, Marcel; Baffelli, Simone; Hajnsek, Irena
Ground-based radar systems are a valuable tool for monitoring natural and built up environment, as they have high flexibility in terms of location, acquisition time, and revisit time. Some systems can provide both interferometric data, which can be used to extract information about scatterer displacement, and also polarimetric data, which gives information about different scattering mechanisms present within a single resolution cell. A new portable Ku-band polarimetric radar interferometer KAPRI – a terrestrial system based on the GPRI (Gamma portable radar interferometer) – has been recently calibrated for monostatic use [1]. KAPRI is also capable of performing acquisitions in the bistatic configuration, in which the transmitting and (one or more) receiving antennas are spatially separated. This configuration has several promising applications – for example, as opposed to the monostatic configuration in which only displacement along the range direction of the radar can be detected, bistatic systems can measure displacements along multiple directions, potentially permitting estimation of two- or threedimensional displacement fields [2]. Furthermore, bistatic measurements can also provide access to more polarimetric parameters of the target, since the Sinclair scattering matrix is no longer necessarily symmetrical (SHV≠SVH) [3], and provide the opportunity to distinguish several scattering mechanisms, which are not the same as in the monostatic case. Several technical challenges have to be overcome, e.g. synchronization of multiple received signals, processing of larger volumes of data, and ensuring that all receivers are correctly and uniformly calibrated. Furthermore, interpretation of data acquired using KAPRI is made more difficult by the fact that, while scattering behaviour of natural surfaces in bands commonly used by space- and airborne SAR systems (mainly in L-, C-, & X-bands) is relatively well known, shorter wavelengths (such as the Ku-band in which KAPRI is operating) are much less studied, since polarimetric devices operating in these bands are still rare. REFERENCES [1] S. Baffelli, O. Frey, C. Werner, and I. Hajnsek, “Polarimetric Calibration of the Ku-Band Advanced Polarimetric Radar Interferometer,” IEEE Transactions on Geoscience and Remote Sensing, vol. 56, no. 4, pp. 2295–2311, 2018. [2] M. Pieraccini, and L. Miccinesi, “Bistatic ArcSAR,” 2018 2nd URSI Atlantic Radio Science Meeting (AT-RASC), pp. 1-4, 2018. [3] J.-S. Lee, and E. Pottier. “Polarimetric radar imaging: from basics to applications,” CRC press, pp. 74-80, 2009.
Exploring polarimetric features of land surface in wetlands and agricultural ecosystems using quad-pol C-band and X-band SAR.
Fitrzyk, Magdalena; Mróz, Marek; Mleczko, Magdalena; Patruno, Jolanda; Delgado Blasco, Jose Manuel
In the context of a larger study on the applicability of multi-frequency and multi-polarimetric data in land cover and vegetation mapping the authors have performed an experiment using quasi-synchroneous observations in C-band and X-band SAR to observe land cover on wetlands and in the agricultural ecosystems. The main purpose of this experiment was: i) to explore polarimetric features extracted from various polarimetric decompositions for C-band routine quad-pol observations with Radarsat-2 and X-band quad-pol bistatic observations with TerraSAR-X obtained in the Science Phase and ii) to find their relationship with diverse land surface characteristics. Results from polarimetric decompositions were then correlated with Sentinel-1 C-band dual pol observations available for the same period and spatial extent. Polarimetric features extracted from both quad-pol C-band and X-band observations didn't reveal substantial differences in the identification of various vegetation associations within the herbaceous wetland. On the other hand some types of land cover and agricultural crops were better distinguishable through decompositions than using dual-pol dataset, particularly comparing to VV/VH bands pair. In the case of short, sparse vegetation covering flat, sandy and relatively smooth parcels forming the wastelands, non-usable for agricultural purposes, there was no significant advantage of the quad-pol configuration over dual-pol one in their separation from other land cover types like grasslands and meadows. Additionally to polarimetric decompositions of quad-pol data acquired by TanDEM-X bistatic formation, the interferometric coherence between channels was also explored. The results achieved in this “3-day – 3-sensor experiment” are discussed in details in the perspective of further wetland and agricultural landscape mapping with available SAR sensors over the specific AOI.
Use of polarimetric information for improving object detection for fully polarimetric SAR
Connetable, Paul; Skriver, Henning; Nielsen, Allan Aasbjerg
The use of fully polarimetric SAR is well known to improve both ground cover and object classification. It indeed provides very useful additional information about the scattering mechanism, such as the number of bounces. In particular, this polarimetric information can be used for man-made object detection, by highlighting specific and recurrent scattering behaviors shared by objects to be detected. An extensive amount of research in the target detection field has been published, and has in a first time focused primarily on using the received power, as in [2]. More recently, several publications have used fully polarimetric data aimed at target detection, as for example [4] and [5]. Their conclusions underline that they can, in particular, and unlike the previous kinds of algorithms, detect targets that do not have a radar cross section higher than their surroundings, as it can happen for example for a vehicle under tree cover. My objective is to use the polarimetric information in the best possible way to improve target detection using a polarimetric SAR. For this purpose, it is of special interest to try a maximum of polarimetric decompositions, as for example the very known [1] or [3], and compare how good the parameters they provide are for target detection. Their individual contribution can be compared by examining the contrast they offer between a target and the clutter, by both increasing the target to clutter ratio and decreasing the clutter variance. This approach has been further generalized to more types of clutter by evaluating for several man-made targets the distribution of their different normalized polarimetric parameters, and comparing it to the same distributions for different types of clutter. This first study has yielded interesting results for obtained contrasts which will improve detection capabilities. Furthermore, the next step of the study is to perform a feature selection, in order to use simultaneously information conveyed by several different polarimetric decompositions, in an optimized way. It is expected that this approach can improve even further the target detection capabilities, by making use of a maximum of known information. References [1] Cloude, S. R., and Pottier, E. An entropy based classification scheme for land applications of polarimetric SAR. IEEE Transactions on Geoscience and Remote Sensing 35, 1 (1997), 68–78. [2] Di Bisceglie, M., and Galdi, C. CFAR detection of extended objects in high-resolution sar images. IEEE Transactions on geoscience and remote sensing 43, 4 (2005), 833–843. [3] Freeman, A., and Durden, S. L. A three-component scattering model for polarimetric SAR data. IEEE Transactions on Geoscience and Remote Sensing 36 , 3 (1998), 963–973. [4] Marino, A., Cloude, S., and Woodhouse, I. A polarimetric target detector using the huynen fork. IEEE Transactions on Geoscience and Remote Sensing 48 , 5 (2010), 2357–2366. [5] Marino, A., Cloude, S. R., and Woodhouse, I. H. Detecting depolarized targets using a new geometrical perturbation filter. IEEE Transactions on Geoscience and Remote Sensing 50 , 10 (2012), 3787–3799.
Forest above ground biomass estimation approach based on multi-dimensional SAR data
Zhao, Lei; Chen, Erxue; Li, Zengyuan; Zhang, Wangfei
At present, most of the forest above biomass estimation (AGB) methods are using the traditional single-dimensional SAR system (1-D at frequency, polarization, or interference dimensional), which has the problems of low precision and poor applicability. In this paper, we studied the synergetic estimation approach of forest AGB based on the multi-dimensional SAR system (at least 2-D at frequency, polarization, or interference dimensional). Based on the air-borne CASMSAR system of China, we acquired the multi-dimensional SAR data (dual antenna X-band InSAR and P-band PolSAR data) covering the boreal forest. Firstly, the high-resolution DEM and CHM (forest height) data of the experimental area was obtained by a filtering method from the X-InSAR data. Then, the terrain correction of P-PolSAR data and X-InSAR coherence was completed using the filtered DEM. Finally, forest AGB was estimated based on the multi-dimensional SAR features (X-/P-band backscatter coefficient; X-InSAR coherence) that after the terrain correction and the filtered forest height information. The experimental results showed that the combined multi-dimensional SAR features can obtain higher estimation accuracy than the single-dimensional SAR features.
Investigating the Dual-pol Behavior for the Derivation of Sea-Ice Surface Topographic Height from TanDEM-X Interferometric SAR Data
Huang, Lanqing; Hajnsek, Irena
Antarctica owns the largest ice sheet in the world. The topography of ice is dominated by ice ridges, shear zones, rubble fields, and hummocks, leading to an intermittent change of the ice topography. Various instruments have been employed to measure the sea ice topography, including laser altimeter measurements, helicopter-borne stereo camera data with photogrammetric techniques, and ground-based laser systems. However, the major limitation of these measurements is their limited spatial coverage. Synthetic aperture radar (SAR) has become an invaluable asset for monitoring polar regions, since it is capable for providing continuous all-weather day/night imagery. The technique of interferometric SAR (InSAR) is employed to interpret topographic information of the earth’s surface. However, due to the dynamic nature of sea ice and the lacking of single-pass interferometric data, the study of the InSAR sea-ice surface topography retrieval is limited. TanDEM-X is a synchronized SAR satellite formation of two satellites and can be regarded as a single-pass SAR interferometer [1]. With its single pass nature combined with the flexibility with respect to both spatial and temporal baseline, the retrieval of sea-ice surface topography in a relatively short-time window can be possible [2]. Hence, for the first time, TanDEM-X offers the chance to study the feasibility of retrieving sea-ice surface topography from space-borne radar. Very recently, focusing on fast multiyear ice with no motion, the sea-ice topographic height derived from TanDEM-X data is validated with results from laser profiler and photogrammetry [2]. Dierking et al. [3] investigated the potential of InSAR measurements for sea-ice surface topography with several satellite configurations and radar frequencies. This paper also demonstrates two examples of the ice surface topography profiles derived from TanDEM data. The examples indicate the need to systematically study the influence of open water surface currents and ice drift on the retrieval of sea-ice surface topography [3]. Aiming at sea-ice surface topography retrieval, several factors which contaminate the sea-ice height retrieval need to be considered and carefully tuned. First, a coherence map can be used to mask the open water areas in the sea-ice surface, since wind-driven surface current on open water areas within the ice cover may generate a phase shift. Besides, the phase difference induced by the spatial variations of the ice drift and rotations of the pack ice need to be removed. Moreover, the phase noise caused by surface and volume scattering effects should be qualified. Since TanDEM-X is an X band radar, volume decorrelation only need to be taken into account for low-salinity mulit-year ice. In this case, the classification maps separating multi-year, first year, and thin ice obtained from SAR intensity images can be helpful for judging. The objective is to propose a well-tuned model which takes all above factors into account and achieve accurate sea-ice surface topography retrieval. REFERENCES [1] G. Krieger et al., “TanDEM-X: A satellite formation for high-resolution SAR interferometry,” IEEE Trans. Geosci. Remote Sens., vol. 45, no. 11, pp. 3317–3341, Nov. 2007. [2] T. G. Yitayew et al., "Validation of Sea-Ice Topographic Heights Derived From TanDEM-X Interferometric SAR Data With Results From Laser Profiler and Photogrammetry,” IEEE Trans. Geosci. Remote Sens., vol. 56, no. 11, pp. 6504–6520, Nov. 2018. [3] W. Dierking, O. Lang, and T. Busche, “Sea ice local surface topography from single-pass satellite InSAR measurements: A feasibility study,” Cryosphere, vol. 11, no. 4, p. 1967–1985, 2017, doi: 105194/tc-2017-40.
Vegetation Sesonal Monitoring with Sentinel-1 Dual Polarization Data
FRISON, Pierre-Louis; FRUNEAU, Bénédicte; KMIHA, Syrine; DUFRENE, Eric; SOUDANI, Kamel; KOLECK, Thierry; VILLARD, Ludovic; LE TOAN, Thuy; MOUGIN, Eric; RUDANT, Jean-Paul
The asset of high temporal frequency of acquisitions for the monitoring of surface parameters has already been pointed out with data acquired by spaceborne scatterometers on board ERS or Metop satellites, for example [1-4]. The acquisitions realized at 1- to 5-day period are particularly well suited for the monitoring of seasonnal variations of land surfaces. The associated coarse spatial resolution (10 to 50 km) dedicates their use to regional or global scales. The Sentinel-1 mission allows to acquire radar data every 12 days from the period extending from March 2015 to August 2016, when only Sentinel-1A operated alone. This period is reduced to 6 days since September 2016 by the combination of Sentinel-1B acquisitions. Consequently, it allows fo²r the first time to assess the potential of radar data acquired with this revisiting time associated to a spatial resolution of about 20 m. In addition to the radar intensity, data acquired by the CSAR (C-band Synthetic Aperture Radar) onboard Sentinel-1 allow to access to he phase information. Consequently, interferometric products between two consecutive aquisitions can be generated, such as the interferometric coherence ρ. In this study we analyze temporal profiles of radar backscattering coefficient σ0 as well as coherence |ρ| over a temperate forest and an agricultural area. Past studies have shown the potential of coherence for vegetation discrimination [5-7]. The two study sites are located in France: one is the Fontainebleau Forest, near Paris, and the agricultural area is located near Toulouse. It appears that the polarization ratio σ0_vv / σ0_vh shows a marked seasonality (with a yearly amplitue of about 3 dB) over stands composed of deciduous trees (oaks and beeches) and a close correspondance with the NDVI that is derived from LANDSAT 8 OLI data. By contrast, no seasonal variations is observed over stands composed of sempervirent species (pines and firns). This confirms the sensitivity of C band to leaves and stems. The interferometric coherence temporal behaviour appears low and constant, near the noise level (aout 0.2) at both polarizations. The most spectacular temporal signatures are observed over agricultural areas. Both σ0_vv / σ0_vh and coherence exhibit striking variations (up to 6 dB and 0.6 resp.). This variations are in close correspondence with field works and vegetation development. This open new capabilities for vegetation monitoring with SAR sensors for the future. Electromagnetic simulations are presently performed for a deeper analysis of these spectacular observations. REFERENCES [1] Frison P.L., Mougin E., «Use of ERS-1 Wind scatterometer data over land surfaces», IEEE Trans. Geosci. Remote Sensing, 34(2), p. 550-560, 1996. [2] Jarlan L., Mougin E., Frison P.L., Mazzega P., Hiernaux P., «Analysis of ERS Wind Scatterometer Time Series over Sahel ( Mali )», Remote Sensing of Environment, 81, p. 404–415, 2002. [3] Magagi R., Kerr Y., «Retrieval of Soil Moisture and Vegetation Characteristics by Use of ERS-1 Wind Scatterometer over Arid and Semi-Arid Areas», Journal of Hydrology, 189(1-4), p. 361–384, 1997. [4] Woodhouse I.H., Hoekman D.H., «Determining Land-Surface Parameters from the ERS Wind Scatterometer», IEEE Transactions on Geoscience and Remote Sensing, 38(1), p. 126–140, 2000. [5] Schmullius, C., J. Baker, H. Balzter, M. Davidson, D. Gaveau, M. Gluck, A. Holz, T. LeToan, A. Luckman, U. Marschalk, S. Nilsson, S. Quegan, 2001: SIBERIA – SAR Imagingfor Boreal Ecology and Radar Interferometry Applications, European Commission 4th Framework Project ENV4-CT98-0743 (DG12-EHKN), Final Report, September 2001. Disponible à l’adresse: http://www.siberia1.uni-jena.de/pdf_files/final_report.pdf [6] Santoro M., Askne J., Smith G., Fransson J., 2002, “Stem volume retrieval in boreal forests from ERS-1/2 interferometry”. Rem. Sens. Environ., vol. 81, 19-35. [7] Engdhal M., Pulliainen J., Hallikainen M., 2004, “Boreal forest coherence-based measures of interferometric pair suitability for operational stem volume retrieval”. Geosc. And Rem. Sens. Letters, vol. 1, n°3, 228-231.
Application of Dual-Pol Multi-temporal Sentinel -1 SAR data for Flash Floods Mapping
Mwaniki, Mercy Wanjiru
Flash Floods are increasingly becoming one of the most devastating natural hazards in many parts of the world after storms and earthquakes (Wilby and Keenan, 2012). Their frequency is depended on weather events which are increasingly impacted by climate variability and climate change. Recent weather patterns in Kenya are characterised by short intense heavy rainfall events resulting in floods which have adverse effects on both rural and urban livelihoods. The motivation for this study is the need to rapidly map areas affected by flash floods and hence facilitate rapid risk and impact assessment as well as guide planning for the city of Nairobi. Consequently, dual-pol Sentinel-1 data were explored as they have been shown to improve flood mapping owing to the double bounce enhancement which increases backscatter from flooded areas (Pulvirenti et al., 2018). The methodology involved processing multi-temporal dual-pol Sentinel-1 data taken prior and post the flood dates by thresholding the individual polarimetric backscatter images of processed dual-pol images to map out water bodies. The results were investigated for the influence of vegetation and urban areas after which a three time-series stack image of the individual time series images were co-registered to map out areas covered by water during the various epochs. The results revealed enhanced capability of the VV polarimetric band in mapping out flooded areas compared to the VH polarimetric band; a characteristic feature of the VV polarimetric band due to less influence of the vegetation. In addition, the results showed the footprints of areas affected by floods for the available image acquisitions prior and post flood date event.
Investigating the implementation issues of speckle filters for Single look Hybrid PolSAR data
Sharma, Rakesh (1); Kumar, Ajeet (2); Panigrahi, Rajib (2) - 1: NIT Hamirpur, India; 2: IIT Roorkee, India
Speckle is a stochastic and correlated noise (or interference pattern) in synthetic aperture radar (SAR) images. It arises due to coherent re-radiation from several scatterers in the radar resolution cell. Despeckling is the process of eliminating speckle noise from the SAR information signal. This is an essential and the foremost pre-processing step for SAR image processing in the field of remote sensing. Speckle noise is critical due the properties like its signal dependence and multiplicative nature. It degrades the image quality and information perception of high resolution images. The visual appearance and polarimetric properties of data are distorted and make information extraction and estimation difficult from the acquired data-set. In the recent past, the patch based non local means (NLM) filtering methods for speckle noise reduction for SAR and polarimetric SAR (PolSAR) have gained popularity because of their edge preserving capabilities along with efficient suppression of the speckle noise. The noise smoothening based on NLM was proposed by Buades et al. in 2005, which have become popular approach for despeckling of PolSAR data in the recent times. The patch based NLM filter for SAR and PolSAR data have best performance in terms of speckle reduction and preservation of sharp details. The hybrid-pol configuration presents essential advantages as compared to full PolSAR configurations. The full PolSAR systems suffer from limitations of double pulse repetition frequency, double average transmit power, half swath coverage and higher data volume. The advancements in processing of hybrid-pol data are carried out targeting classification performance close to that of full-pol configurations. The processing of hybrid-pol data is in the development phase and different decomposition \& classification methods are being developed. The decomposition methods for hybrid PolSAR systems is based on Stokes data format like $m-\delta$, $m-\alpha$ and $m-\chi$. The speckle filtering of hybrid-pol data in Stokes format is absent in the literature. In this paper, the patch based non local means speckle reduction algorithms with different similarity and dissimilarity measures based on Stokes format is implemented for hybrid PolSAR systems. The possibilities of using patch based NLM filter for hybrid-pol data is also explored by evaluating the performance of filter for speckle reduction, preservation of radiometric property and classification results. The proposed filter is applied to synthesized hybrid-pol single look data from Radarsat-2 (San Fransisco bay region) and real hybrid-pol data from ISRO's RISAT -1 (over Mumbai coastal area, India). The filter is compared based on visual results, ENL, mean and standard deviation for the ratio image (noisy by filtered image), and classification accuracy with and without using NLM filter to state of art speckle filters. Hence, apart from measuring the extent of speckle filtering which is measured by ENL, unbiased estimation is measured by mean and standard deviation of the ratio image. Finally, the effect of speckle filtering on landcover classification is also measured by classification accuracy.
POLARIMETRIC SAR INTERFEROMETRY SATELLITE SYSTEM AND PARAMETER OPTIMIZATION DESIGN
Xu, Liying; Wei, Li; Fang, Dong; Wang, Haitao; Tong, QingWei; Yu, Yingjun; Chen, JunLi - Shanghai Institute of Satellite Engineering, Shanghai, 200240, China
PolInSAR has important application value in many fields such as biomass estimation, terrain target classification, high-precision DEM extraction, and acquisition of micro-topography change detection. Now it has already achieved achievability with the key technology research’s completion of the low frequency SAR. Because the applications of PolInSAR are different, there are special requirements for system parameters. The current system parameter design of PolInSAR is simply the method using PolSAR or InSAR. In this paper, the parameter design method of PolInSAR system is proposed based on the application requirements. The key parameters, including baseline, polarization isolation and channel imbalance, of the PolInSAR system are given. The design method and examples will provide a theoretical basis for the development of PolInSAR satellite system.
09:30 - 10:10
09:30 -
TanDEM-X tomography: Perspectives from experiments over multi-species Indian tropical forest
Khati, Unmesh (1); Lavalle, Marco (2); Singh, Gulab (1) - 1: CSRE, Indian Institute of Technology Bombay, Mumbai, India; 2: Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA, USA
TerraSAR-X and TanDEM-X form the first space-borne single-pass satellite system capable of acquiring fully-polarimetric SAR data with zero- or near-zero temporal baseline. These satellite systems have been successfully utilized for accurate forest height and forest above-ground biomass (AGB) estimation. In our earlier work, we explored the potiential of TanDEM-X data for forest height estimation over Indian tropical forest and compared them with ALOS-2/PALSAR-2 and RadarSAT-2 PolInSAR inversion. SAR Tomography (SAR) takes advantage of multiple SAR acquisitions to provide 3-D vertical structure of the imaged target. Over the last decade, space-borne tomography has progressed with various researchers exploring the potential over urban- and forest-targets. Recently, potential of TanDEM-X for forest 3-D structure estimation has been explored over Boreal and Tropical forests using TomoSAR. However, an in-depth analysis of multi-polarization SAR tomography over tropical forests is lacking. Tropical forests form one of the most complex ecosystems with a high diversity of species, structure and phenology. This paper utilizes 18 TerraSAR-X/TanDEM-X acquisitions in multi-polarimetric configuration acquired over Haldwani forest range in sub-Himalayan North India. The data sets are processed using Capon beamforming to generate TanDEM tomograms over this tropical Indian forest. The study test site selected is a managed forest with uni-species distribution among its forest compartments. The TomoSAR data is processed over four major species – Teak (Tectona grandis), Eucalyptus sp., Poplar sp. and Gutel (Trema orientalis). These species have distinct forest structural characteristics - leaf- and canopy-density, leaf-size, canopy vertical structure and phenology. The TanDEM TomoSAR data is analyzed over these different species in multiple polarizations. Field survey provides in situ information at 100 different plot locations with the forest stand height and forest AGB (above-ground biomass) measured. The tomograms and vertical profiles are generated over multiple plots for each species and analysed in detail. The TanDEM tomograms show distinct 3-D structure across the species. It was observed that at X-band, the canopy gaps and leaf-density play a crucial role for microwaves penetration through ground. The TomoSAR vertical profiles for the four major species can be summarized as: 1) the high canopy- and leaf-density species of Gutel have distinct tomograms which depict a strong scattering from the canopy and negligible penetration through the ground. 2) Teak, with high leaf-density but with canopy-gaps lead to tomograms with a small contribution from the ground and the coherence backscatter spread through the canopy. Similarly for low canopy- and leaf-dense 3) Eucalyptus and 4) Poplar plantations, the dominant scatterers are the canopy and ground with a good agreement with field observations. Further, the TanDEM tomogram is utilized for estimation of forest AGB using simple linear regression models. It is assumed that the TomoSAR backscatter layers correspond with the biomass content at that particular layer of the forest. This provides a relation between the AGB measured and the TomoSAR vertical backscatter along different elevations and in different polarization. The best AGB estimation is observed in the HH-pol tomogram at 27 m height with a correlation coefficient r = 0.76 and RMSE of 50 t/ha or %RMSE of 30. In conclusion, this paper provides a detailed analysis of TanDEM-X tomograms over a multi-species Indian tropical forest. It is one of the first studies to analyse the potential of single-baseline shorter wavelength SAR data for tropical forest tomography. The results are highly encouraging with the potential to capture subtle variations in the structure across species. One major limitation is posed by the high extinction and inability to accurately estimate the ground. This can be addressed in the proposed TanDEM-L mission and the upcoming BIOMASS with a dedicated tomographic observation phase.
09:50 -
Ground Topography Below Tropical Forests through Polarimetric SAR Tomography
Mariotti d'Alessandro, Mauro; Tebaldini, Stefano - Politecnico di Milano
Accurate topographic maps are requested for both commercial and scientific uses. In the past decades, SAR interferometry provided excellent elevation maps on bare soils. The high accuracy and good revisit time also allowed the observation of the temporal evolution of the Earth's topography. However, when complex media are considered, InSAR measurements are not sufficient to clearly locate the ground level; typically, it happens when vegetation or trees cover the ground surface. In this case the interferometric phase refers to an elevation placed somewhere between ground and tree top depending on forest density and wavelength; the shorter the wavelength the smaller the penetration capability. As an example, the excellent DEM returned by the TanDEM-X mission refers to the first centimeters of the forest top height over most of the forested areas. P-band signals offer a significantly better penetration even through the tropical forest considered in this work. Still, the phase center is not exactly at the ground level: the elevation bias changes with ground slope and incidence angle making it difficult to model it. These difficulties arise from the attempt of describing such a complicated scenario with only one observable: the interferometric coherence. In order to cope with these scenarios a larger number of observables is needed. Many coherent images of the same area provide a whole set of interferometric coherences to be jointly processed and allow a different set of processing tool to be exploited; SAR tomography can be carried out. The range of possibilities is further widened by multi polarimetric acquisitions; a multi-polarimetric image stack provides physical characterization of scattering mechanisms that can be separated in height. Also, by jointly considering the polarimetric and the interferometric features, sets of elementary scatterers can be recognized thus moving from scatterers to scattering mechanisms. In a forest scenario two scattering mechanisms are often enough to interpret the observations: the ground scattering mechanism and the volume (that is the vegetation layer). The interferometric analysis of the ground scattering mechanism provided very good estimates of the ground topography under many boreal forests. This work aims to extend the range of availability of this class of processing techniques to include tropical forests as well. An analysis of the elevation map provided by InSAR processing is presented with focus on the effects connected to geometry of acquisition and terrain topography. The DTM obtained by means of PolTomoSAR processing techniques are then shown. Their accuracy is assessed by comparing with LiDAR estimates. Results are based on the tomographic stacks acquired by DLR in the framework of the AfriSAR campaign. They are made of several (about ten) multi-polarimetric SLC images acquired on very dense tropical forest in Africa as a support activity for the forthcoming ESA BIOMASS mission. The three sites cover different kinds of forest composition with tree top height reaching 60m and biomass amount up to 600T/Ha. The data for comparison comes from two different acquisitions carried out in the framework of the 2016 NASA AfriSAR campaign. The Mabounie site has been imaged by the LVIS LiDAR system carried by the LaRC (Langley Flight Research Center) B-200 aircraft; the Mondah and Rabi sites by the Riegl VQ480U sensor carried by the EC 135 helicopter. The former acquisition is characterized by a large footprint (about 20m) whereas the latter by fine beam pulses (about 10cm). The results here presented show that PolTomoSAR at P-band provides topographic maps under tropical forest with an accuracy at least comparable to LiDAR systems. Furthermore, the difference between tomographic and laser estimates changes significantly in case of large footprint LiDAR or fine beam. This suggests that the difference between the two should not be considered as error associated with tomographic processing but rather as difference between two reliable estimates.
10:10 - 10:40
10:40 - 12:00
10:40 -
SARSIM and SARSIM+: data-bases for the development of SAR Tomography in forestry applications
Tebaldini, Stefano (1); Mariotti d'Alessandro, Mauro (1); Pardini, Matteo (2); Ferro-Famil, Laurent (3); Huang, Yue (3); Papathanassiou, Kostas (2); Ulander, Lars (4); Blomberg, Erik (4); Wasik, Valentine (5); Dupuis, Xavier (5); Lavalle, Marco (6); Hensley, Scott (6); Scipal, Klaus (7); Albinet, Clement (7) - 1: Politecnico di Milano, Italy; 2: DLR - Department Radar Concepts; 3: Université de Rennes 1; 4: Chalmers University of Technology; 5: ONERA; 6: NASA JPL; 7: European Space Agency
The SARSIM and SARSIM+ data-bases have been created in the frame of the ESA study “L- and P-band SAR Tomography Synergies Consolidation Study” with the aim of constituting a complete reference data-set for current and future researches on the application of SAR Tomography for the remote sensing of boreal, temperate, and tropical forests at P- and L-Band. The intended users of SARSIM and SARSIM+ are: Signal processing researchers interested in developing new tomographic processing approaches without having to implement any pre-processing operation; Remote sensing and ecology researchers interested in using tomographic data for the retrieval of forest parameters; Researchers interested in developing 3D forest scattering models; Graduate and Ph.D. students willing to learn SAR tomographic processing and/or experiment with tomographic data. The SARSIM data-base includes SAR Single-Look-Complex (SLR) image stacks from previous campaigns, as well as simulated spaceborne data derived from campaign data. All data in SARSIM have been accurately phase calibrated to ensure accurate tomographic focusing. Accordingly, future users will be able to use these data without having to implement any pre-processing operation. SARSIM+ includes 3D tomographic voxels derived from the data-sets in SARSIM. These products are provided in ground coordinates with respect to a reference DTM, so as to facilitate comparison with Lidar or in-situ data. All data have been saved in TIFF format. DEMO Matlab codes are provided to show new users how to load the data, visualize them, and implement basic tomographic processing. The aim of this talk is to present the main features of SARSIM and SARSIM+, the included data-sets, and discuss example of applications for the retrieval of forest height and AGB.
11:00 -
The role of polarimetry on the estimation of forest structure from SAR tomography
Cazcarra-Bes, Victor; Pardini, Matteo; Tello-Alonso, Marivi; Papathanassiou, Konstantinos - German Aerospace center (DLR), Germany
Forest structure can be understood as the description of the distribution of trees and tree elements on a three dimensional (3-D) space. It is a key parameter of forest ecosystems and allows to fully characterizing the forest [1]. Moreover, forest structure is essential to better assess biomass and forest productivity. Therefore, extracting 3-D information of forest is the first step in order to further obtain forest structural products. From the remote sensing point of view, one way to obtain 3-D information from the forest is using synthetic aperture radar (SAR) tomography techniques [2]. SAR tomography extends the concept of synthetic aperture of the classical two dimensional SAR imaging by combining several SAR images over the same area with slightly different position. This concept, together with the use of a low frequency band (L- / P-band) capable to penetrate trough the canopy until the ground under forest, makes possible to obtain a 3-D radar reflectivity of the different elements (trees, brunches) of the forests. Therefore, the use of SAR tomography allows extracting 3-D information of the forest. However, the physical interpretation of the radar reflectivity to forest structure information is not straightforward. The results from tomography depend not only on the physical distribution of the elements, but also on the dielectric properties of them as well as system parameters like the tomographic algorithm, the frequency or the polarizations used [6]. Despite these difficulties, a new approach has been recently presented [3] as an attempt to obtain 3-D forest structure with an ecological interpretation from SAR tomography. This concept has been also applied to monitor changes of forest structure on a temperate forest [4] and to get forest structure information on a tropical scenario [5]. As already mentioned, the use of different polarizations has an effect on the tomographic result and, as a consequence, on the resulting forest structure from it. In this context, the goal of this paper is to analyse the role of polarimetry on the estimation of forest structure as defined in [3]. More in detail, from one side we will show the limitations of using one single polarimetric channel instead of a fully polarimetric acquisition on the estimation of forest structure. On the other side, we will consider a case where only one polarization is available, to assess the differences between the co-pol and cross-pol channels. In order to do so, we will use a multi-baseline and fully polarimetric SAR dataset acquired in a repeat pass mode at L-band by the German Aerospace Center (DLR) F-SAR system. The test site under study is a temperate managed forest located in Traunstein in the south-east of Germany. Moreover, to validate and discuss the 3-D forest structure results obtained with tomographic SAR data, a data set of more than 16000 trees and a high resolution Lidar over the same area are available. References [1] T. A. Spies, “Forest structure: a key to the ecosystem”, Northwest Science, vol. 72, no. 2, pp. 34–36, 1998 [2] A. Reigber and A. Moreira. “First demonstration of airborne SAR tomography using multibaseline L-band data.” IEEE Trans. Geosci. Remote Sens., Vol. 38, No. 5, pp. 2142–4152, Sep. 2000. [3] M. Tello, V. Cazcarra-Bes, M. Pardini, and K. Papathanassiou, “Forest Structure Characterization From SAR Tomography at L-Band,” IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, no. 99, pp. 1–13, 2018. [4] V. Cazcarra-Bes, M. Tello-Alonso, R. Fischer, M. Heym, and K. Papathanassiou, “Monitoring of forest structure dynamics by means of L-band SAR tomography,” Remote Sensing, vol. 9, no. 12, p. 1229, 2017. [5] M. Pardini, M. Tello, V. Cazcarra-Bes, K. P. Papathanassiou, and I. Hajnsek, “L-and P-Band 3-D SAR Reflectivity Profiles Versus Lidar Waveforms: The AfriSAR Case,” IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, no. 99, pp. 1–16, 2018. [6] O. Frey and E. Meier, “Analyzing tomographic SAR data of a forest with respect to frequency, polarization, and focusing technique,” IEEE Transactions on Geoscience and Remote Sensing, vol. 49, no. 10, pp. 3648–3659, 2011.
11:20 -
Analysis of Temporal Decorrelation Effects on Point Spread Function of 3-D Tomo Beamforming
Lombardini, Fabrizio (1); Aghababaee, Hossein (2); Schirinzi, Gilda (2) - 1: University of Pisa, Italy; 2: University of Napoli Parthenope, Italy
After about two decades of research activities and experiments, 3-D SAR Tomography [1], stemming from multibaseline SAR Interferometry [2] to get resolution capabilities in the height dimension for remote sensing of complex scenarios, has matured tending to the operational level [3-8]. Beyond applications to layover solution in height mapping of urban areas [8], the other most investigated remote sensing application is 3-D imaging of forest layers [1,3,5,7,9], especially for biomass monitoring, and dedicated spaceborne missions have been studied or are under development to this goal [3]. However, the main investigated, experimented, and adopted 3-D Tomography focusing algorithms, that are based on array processing of the multibaseline complex data obtained by multiple passes of a standard SAR system for height beam forming and steering [1,7,9], are often affected by temporal coherence loss of the (natural) scatterers [2,10]; this is well known to produce defocusing and blurring effects in the height imaging [1,9,10]. Beyond the possibility to consider resorting to special decorrelation-robust focusing 4D (3D+Time) algorithms [9] applicable for some baseline-time acquisition patterns, or to consider dedicated advanced SAR system configurations, the issue of temporal decorrelation still constitutes a particular criticality in the development of operational spaceborne missions for forest Tomography, such as the ESA Biomass mission [3]. Oddily enough, during the maturation of SAR Tomography, beyond some practice and indications got from experimental observations and simulated analyses, practically no effort has been dedicated to the development of an analytical theory to study and understand effects of temporal decorrelation on 3-D imaging, while a couple of papers have been dedicated to analyze effects of phase miscalibration [4,6], and recently seasonal and weather effects have been tackled. In this context, in this work an analytical theory is presented of the statistical behaviour of the height Point Spread Function (PSF) of 3-D SAR Tomography, with reference to the Fourier beamforming focusing algorithm [1,6,10], in presence of temporal decorrelation; the theory is based on first closed form derivations [10] obtained at the beginning of the 3-D SAR Tomography area, and is expanded also presenting general yet representative case studies, and more specific examples. It is expected that this analytical statistical characterization of the PSF of Fourier beamforming Tomography for a general partially temporal coherent scatterer can be useful, being the PSF the first main indicator of imaging quality for any image formation technique, and Fourier beamforming being one of the most diffused algorithms for array processing-based SAR Tomography. The presented analytical derivations in particular, given a typical expected or representative temporal decorrelation model [2,10], allow to easily quantify the average behaviour of the defocused and blurred PSF shape, and to get corresponding insights on different trends of PSF behaviour for different temporal decorrelation scales and baseline acquisition patterns versus time; in an operational context, this paves the way to a handy imaging quality prediction for 3-D Tomography oriented system feasibility studies, mission planning, and optimization, accounting in a detailed quantitative manner for temporal decorrelation and complementing the phase miscalibration oriented developments [4,6]. Specifically, the analytical statistical PSF characterization is particularized for both typical short- and long-term decorrelation models [2,10], and normalized general-purpose yet well representative case studies are reported for both regular baseline acquisitions in time and scattered baseline-time acquisition patterns, showing the corresponding average 3-D Tomography PSF shape for varying decorrelation degrees, that is different in terms of mainlobe and sidelobes behaviours, respectively; related insights are given. Analysis of the statistical dispersion of the PSF profile under decorrelation effects is also attacked. Comparison of the analytical derivations with simulations are reported confirming the predictions and illustrating the derived insights. Comparison with examples from real (airborne P-band) data will be possibly included at the conference. Moreover, case studies are shown with parameters mimicking the incoming P-band Biomass system [3], and the future planned NASA-ISRO NISAR system [11] with reference to its L-band channel (more critical for forest spaceborne Tomography that in fact is not currently, at the best of the Authors’ knowledge, a targeted application for this mission); effect of (hypothesized) mission parameter changes are possibly illustrated as well. Hints on the application of this analytical theory of statistical PSF behaviour to an operational oriented imaging quality prediction of 3-D Tomography are finally given. [1] Reigber, A., Moreira, A.: ‘First demonstration of airborne SAR tomography using multibaseline L-band data,’ IEEE Trans. Geosci. Remote Sens., 2000, 38, (5). [2] Rocca, F.: ‘Modeling interferogram stacks,’ IEEE Trans. Geosci. Remote Sens., 2007, 45, (10). [3] Scipal, K., ‘The Biomass mission - ESA'S P-band polarimetric, interferometric SAR mission,’ Proc. IEEE Int. Geosci. Remote Sens. Symp., Fort Worth, Texas, 2017. [4] Pardini M., Lombardini F., Gini F.: ‘The hybrid Cramér-Rao bound on broadside DOA estimation of extended sources in presence of array errors,’ IEEE Trans. Signal Processing, 2008, 56, (4). [5] Pardini, M., Torano Caicoya, A., Kugler, F., Lee, S.K., Hajnsek, I., Papathanassiou, K.: ‘On the estimation of forest vertical structure from multibaseline polarimetric SAR data,’ Proc. IEEE Int. Geosci. Remote Sens. Symp., Munich, 2012. [6] Tebaldini, S.: ‘On the role of phase stability in SAR multibaseline applications,’ IEEE Trans. Geosci. Remote Sens., 2010, 48, (7). [7] Huang, Y., Ferro-Famil, L., Reigber, A.: ‘Under-foliage object imaging using SAR tomography and polarimetric spectral estimators,’ IEEE Trans. Geosci. Remote Sens., 2012, 50, (6). [8] Reale, D., Fornaro, G., Pauciullo, A., Zhu, X., Bamler, R.: ‘Tomographic imaging and monitoring of buildings with very high resolution SAR data,’ IEEE Geosci. Remote Sens. Lett., 2011, 8, (4). [9] Lombardini, F., Cai, F.: ‘Temporal decorrelation-robust SAR tomography,’ IEEE Trans. Geosci. Remote Sens., 2014, 52, (9). [10] Lombardini F., Griffiths H.: ‘Effect of temporal decorrelation on 3D SAR imaging using multiple pass beamforming,’ Proc. IEE-EUREL Meeting on Radar and Sonar Signal Processing, Peebles, UK, 1998. [11] Rosen, P., Hensley, S., Shaffer, S., Edelstein, W., Kim, Y., Kumar, R., Misra, T., Bhan, R., Sagi, R.; ‘The NASA-ISRO SAR (NISAR) mission dual-band radar instrument preliminary design,’ Proc. IEEE Int. Geosci. Remote Sens. Symp., Fort Worth, Texas, 2017.
11:40 -
Sparse Signal Analysis for Full Rank Polarimetric Reconstruction of Coherence Matrix T3
Aghababaee, Hossein (1); Ferro-Famil, Laurent (2); Ferraioli, Giampaolo (1); Schirinzi, Gilda (1) - 1: Università di Napoli Parthenope, Italy; 2: Université de Rennes 1
Synthetic aperture radar (SAR) tomography is a multidimensional signal processing technique based on the use of multi-baseline (MB) SAR data [1]. SAR tomography (TomoSAR) leads to improved imaging and characterization of observed objects by providing the height reflectivity of the imaged scene, allowing the extraction of valuable information for many different applications. In addition, the polarimetric reflectivity of observed scene can be established by means of polarimetric multi-baseline data [2, 3]. Differently from such a polarimetric reflectivity reconstruction techniques, where polarization is employed to improve the synthesizing performance and the discrimination between vertically aligned scatterers, in [4] a full rank reconstruction framework has been proposed that allows fully interpretation of electromagnetic behavior of illuminated objects by reconstruction of the typical coherence matrix T3. The main particularity of such a reconstruction is related to its ability, in which their output T3 can be characterized using classical polarimetric processing algorithms. However, in practice, due to the limitation on baseline designing, full rank reconstruction by typical spectral estimation techniques [4, 5] (like Beamformer, Capon) may bring some quality problems in the target interpretations. In this paper, we consider the approach based on signal sparsity for full rank reconstruction that characterized by the advantage, with respect to classical techniques, of recovering full information from a reduced set of measurements and resolution improvement. Recent works have been addressed the typical tomographic and polarimetric tomographic inversions with the concept of compressed sensing [6, 7], while this paper differs from previous studies in which the proposed method relies on the sparse representation in order to reconstruct full rank polarimetric information of the scatterers at different height levels. More precisely, the method works with second order statistics of polarimetric MB SAR data set, and exploits sparse representation of fully polarimetric information. Once the coherency matrices T3 at different height levels are estimated, all the classical polarimetric processing techniques including H/α/A and lexicographic decompositions [8] can be employed for scene investigation. Here, in order to evaluate the method a stack of 6 fully polarimetric SAR images acquired by ONERA over a tropical forest in French Guyana in the frame of the European Space Agency’s campaign TROPISAR [9] has been employed. In particular, the analysis of sparse signal for full rank polarimetric reconstruction has been investigated. In the following figures, the reconstruction of α angle, span and lexicographic images have been carried out along a specific range line from the selected data set. For the comparison and in order to highlight the efficiency of the proposed method, results from full rank reconstruction by typical Capon algorithm presented in [4] are also included. From the experimental results, resolution improvement archived by sparse reconstruction with respect to the Capon is evident specifically in span and lexicographic images. Along this line, it can be verified that for more pixels the superposition of scatterers from canopy and ground is well addressed using compressive sensing in comparison with Capon. Furthermore, from the lexicographic image by sparse signal assumption in the compressive sensing, the backscattering behavior can be more precisely and simply, interpreted, where the canopy top is dominated by volumetric scattering (green) and ground with double-bounce (red) scattering mechanisms. Additionally, the proposed method significantly diminished the ambiguity of α angle in the interpretation of polarimetric behaviors of scatterers. In the produced α angle by proposed method, double bounce interaction from under foliage and volumetric scattering mechanisms in canopy level are properly identified with α ≈ 90 and α ≈ 45 degrees, respectively, while high ambiguity level can be observed in the results of Capon based reconstruction of α angle. In fact, depart from the high efficiency of sparse full rank reconstruction, the proposed model is able to cope with some possible ambiguities by reconstruction of coherence matrix T3 simultaneously for all heights in the defined elevation space. Theoretical formulations and details of implementation together with more experimental results from other typical polarimetric tools will be reported at the conferences.
12:00 - 13:00
13:00 - 13:40
13:00 -
On the role of ground/volume decomposition for AGB retrieval
Banda, Francesco (1); Mariotti d'Alessandro, Mauro (2); Tebaldini, Stefano (2); Giudici, Davide (1) - 1: aresys, Italy; 2: Polimi, Italy
In this work we investigate the role of volume scattering obtained from SAR tomography (TomoSAR) [3] and interferometry in retrieving AGB (Above Ground Biomass). Results here presented originate from the BIOMASS L2 study [5], aimed at defining and implementing the tomographic and interferometric processors of the BIOMASS mission. TomoSAR gives the full 3D of a forest scene by coherently combining multiple acquisitions collected at slightly different baselines. It will be one the main innovations characterizing the BIOMASS mission, allowing the retrieval of unprecedented information by virtue of its capability to single out returns from different layers of the forest canopy. In fact, the latest studies proved that TomoSAR intensity of tropical forests at 30 m canopy layer is strongly correlated to AGB [3]. This result is supported by ecological studies indicating that the fraction of biomass included in the 30 m layer accounts for about 35% to 40% of total AGB over a large range of AGB values [1]. Moreover, TomoSAR gives good rejection of ground scattering (terrain scattering and double bounce), determined by a complex set of factors other than forest biomass and unlikely to be directly related to AGB in an operational context [2]. The interferometric ground notching technique, also presented at this conference, was recently proposed to improve AGB retrieval during the interferometric phase of BIOMASS, when the availability of just three acquisitions per site will prevent using tomography. Ground-notching [4] is obtained by taking the difference between two coregistered, phase-calibrated and ground-steered SLC SAR images, which results in ground scattering being automatically cancelled out. The interferometric height of ambiguity of the image pair determines how scattering from different vegetation layers is weighted in the ground-notched image. From analysis on Paracou data it results that, with a favorable range of height of ambiguities, the resulting IRF emphasizes the main canopy layer, therefore approximating with two images what is done by Tomography with many. In this paper we consider another approach for the separation of ground and volume scattering, i..e: ground/volume decomposition based on single baseline polarimetric and interferometric data. The rationale of this technique is very well known in literature: the assumption of Random Volume over Ground (RVoG) model allows to link the variation of the interferometric coherence with polarization to the physical parameters associated with ground and volume scattering, which include forest height, extinction, as well as to the total backscattered power associated with ground-only and volume-only scattering. Interestingly, the performance of such a decomposition is in principle much less affected by the interferometric height of ambiguity and topographic slopes than ground-notching. In this paper we aim at discussing whether, and to what extent, ground/volume decomposition can provide a valid alternative to ground-notching. To do this, both are tested based on the P-Band collected at the forest site of Paracou, French Guiana, during the TropiSAR campaign, and validated against in-situ AGB measurements in terms of correlation and sensitivity of the retrievals. Quite surprisingly, results indicate that volume-backscattered power as obtained by ground/volume decomposition is almost unsensitive to AGB, notwithstanding different solutions for volume scattering are tested, and lead to conclusion that forest structure actually plays a non-negligible role in AGB retrieval in tropical areas. [1] J.Chave, "Study of structural, successional and spatial patterns in tropical rain forests using TROLL, a spatially explicit forest model", Ecological Modelling., Jun. 1999 [2] M. M. d'Alessandro, S. Tebaldini and F. Rocca, "Phenomenology of Ground Scattering in a Tropical Forest Through Polarimetric Synthetic Aperture Radar Tomography," IEEE Transactions on Geoscience and Remote Sensing, Aug. 2013 [3] D. Ho Tong Minh, T. L. Toan, F. Rocca, S. Tebaldini, M. M. d'Alessandro and L. Villard, "Relating P-Band Synthetic Aperture Radar Tomography to Tropical Forest Biomass," IEEE Transactions on Geoscience and Remote Sensing, Feb.2014. [4] M. M. D'Alessandro, S. Tebaldini, S. Quegan, M. Soja, L. M. H. Ulander, "Interferometric Ground Notching of SAR Images for estimating Forest Above Ground Biomass," IGARSS 2018 [5] F. Banda, D. Giudici, S. Quegan, K. Scipal, "The Retrieval Concept of the Forest BIOMASS Prototype Processor," IGARSS 2018
13:20 -
Polarimetric coherence optimization for 3-D imaging applications
Ferro-Fami, Laurent (1); Huang, Yue (1); Tebaldini, Stefano (2) - 1: University of Rennes 1, France; 2: Politecnico di Milano, Milano, Italy
Various studies have shown that polarimetric coherence optimization, performed over coherent stacks of polarimetric SAR images, might be interpreted differently, depending on the application. It may be employed as an efficient signal extractor, as an optimal projection tool into a space that remains invariant under non-degenerate multivariate transformation ... This paper proposes to demonstrate how polarimetric coherence optimization may be used in the frame of 3-D imaging of volumetric media through polarimetric SAR tomography. In particular it is shown that the proposed optimization may reveal useful to: - significantly simplify the structure of 3-D SAR data acquired in a tomographic configuration - robustify existig estimation techniques - characterize the observed volume though a quasi-analytical model-based tomographic estimation process. The performance of these new techniques are illustrated over SAR data acquired by the DLR at L band over a Bpreal forest, in the frame of the BioSAR II campaign.
13:40 - 14:00
14:00 - 15:20
14:00 -
Biomass mission: Biophysical products retrieval
Quegan, Shaun (1); Giudici, Davide (2); Banda, Francesco (2); Scipal, Klaus (3); Papathanassiou, Kostas (4); Villard, Ludovic (5); Ulander, Lars (6); Soja, Maciej (6); Mariotti d'Alessandro, Mauro (7); Tebaldini, Stefano (7); Le Toan, Thuy (5) - 1: University of Sheffield, England; 2: aresys, Italy; 3: ESA; 4: DLR; 5: Cesbio; 6: Chalmers University; 7: Polimi
The ESA BIOMASS mission will be the 7th Earth Explorer and its primary objective is to measure the above-ground biomass (AGB) in the world’s forests. It is highly innovative, being the first spaceborne P-band SAR mission, fully-polarimetric on all acquisitions, and providing both polarimetric interferometry (Pol-InSAR) and tomographic data. These will be used in synergy to produce Level-2 maps of biomass at 200 m resolution, forest height at 200 m resolution, and severe forest disturbance at 50 m resolution. The mission will also produce tomographic voxels that provide forest backscatter in 3D for each polarization. The current ESA Level-2 (L2) implementation study focuses on defining and implementing the main algorithms for forest parameter retrieval from BIOMASS data. When the study started, the only demonstrated approaches for estimating AGB from P-band data were essentially empirical and needed substantial amounts of reference data, from in situ plots or calibrated airborne lidar. Another drawback of these approaches was that they were not fully representative of the range of conditions of forested environments, so that generality was of concern. The L2 study has made radical improvements in the approach to estimating biomass based on three key elements: establishment of strong evidence that in tropical forests the backscatter from the canopy region 25-35m above the ground is highly correlated with the total AGB (which can be exploited using the full power of tomography ); the development of interferometric ground cancellation to isolate volume scattering; and the development of an approach to invert the model that minimizes the need for reference data. The development of ground cancellation, in particular, has proved to be huge value, since it removes the effects of environmental variability and contributions unrelated to the forest carried in the ground scattering. Forest height methods are well-developed and are currently focused on optimising the algorithms for the BIOMASS case. The forest disturbance algorithm is based on a likelihood ratio test that estimates whether change has occurred in a time series of polarimetric data at a specified level of significance. This paper will discuss the status of the study and our findings from tests of the algorithms using airborne campaign data and simulated data. It will also describe the current activities of the L2 team, which are devoted to systematic validation of the implemented algorithms.
14:20 -
RVoG Model Validity for Forest Height Estimation using SAR L Band over a Heterogeneous Tropical Forest
Pourshamsi, Maryam
Pourshamsi, Maryam (1); Pardini, Matteo (2); Papathanassiou, Kostas (2); Balzter, Heiko (1) - 1: University of Leicester, United Kingdom; 2: DLR (Germany)
In this paper, we investigate the validity of the Random Volume over Ground Model (RVoG) for the estimation of forest height from polarimetric interferometric synthetic aperture radar (PolInSAR) data. First, the RVoG model and PolInSAR processing chains are introduced. Through the review of the forest height estimation from PolInSAR data, the sources of errors on the inversion performance are evaluated. We discuss that a wrong estimation of ground phase might cause a large bias. We therefore, fix the true ground phase estimated from LiDAR for all the available baselines, and then investigate the validity of the RVoG model. The results indicate that the RVoG model is a valid technique for forest height estimation from PolInSAR data. However due to the effect of ambiguity, only small baselines with large height of ambiguity are suitable for the estimation of large heights. The role of the vertical wavenumber and its consideration in the inversion methodology is addressed. The RVoG model validity analysis is demonstrated by means of airborne repeat pass PolInSAR acquisitions with SAR L-band acquired over a heterogeneous forest ranging from short/sparse savannahs to high/dense tropical forest. The estimated forest height is validated against reference height derived from airborne LiDAR acquisitions.
14:40 -
An evaluation of Pol-InSAR complementarities between L- and S-band in forest structure observation
Pardini, Matteo; Cazcarra-Bes, Victor; Papathanassiou, Konstantinos - German Aerospace Center (DLR), Germany
There is a common agreement on the relevance of L-band wavelengths (around 20 cm) for forest observation. For instance, the larger backscatter dynamic range increases the sensitivity to larger biomass gradients with respect to C- and X-band. Furthermore, L-band allows the penetration into and through dense forest canopies in all forest ecosystems, remaining sensitive to canopy structure elements, and possibly leading to more relevant structure estimates than P-band. Finally, the larger temporal stability of the scatterers may allow the implementation of polarimetric interferometric (Pol-InSAR) acquisitions in repeat pass modes. In contrast, S-band applications in the forest domain are rather unexplored. The interest in S-band is being raised by its implementation on a number of space borne platforms, e.g. HJ-1C (China) [1], NovaSAR-S (United Kingdom) [2] and NISAR (United States and India) [3]. Clear potentials of S-band polarimetry have been found for instance for land classification [4], while S-band interferometry has been shown to lead to accurate estimates of digital elevation models for instance in a dual-frequency framework with X-band [5]. Differently from L-band, the shorter wavelength allows the realization of flexible single-pass implementations with relevant interferometric sensitivity even on airborne platforms. Despite the first airborne experiments in [6] and [7], the S-band performance in terms of forest structure characterization from polarimetric interferometric (Pol-InSAR) data has not been systematically evaluated yet. After the first S-band Pol-InSAR data acquisition of [6], larger scale campaigns have been performed by the DLR’s F-SAR airborne platform on a number of test sites. In this context, the purpose of this work is to further investigate the complementarity of L- and S-band for forest structure observation as a consequence of the difference in penetration and sensitivity to sizes of vegetation elements, starting from multibaseline Pol-InSAR data. In particular, the role and importance of polarimetry is compared for the two frequencies. First of all, the spectrum of the ground-to-volume ratios is estimated under the Random-Volume-over-Ground assumptions, and its variability is evaluated as function of terrain slopes and incidence angle. Then, the distributions of the interferometric coherences within the Pol-InSAR coherence region are considered in different forest stands. Finally, the analysis is extended to vertical reflectivity profiles estimated by tomographic techniques. Experiments are carried out by using multibaseline Pol-InSAR data sets at L- and S-band acquired simultaneously by the F-SAR platform over the forest site of Traunstein (South of Germany) with a large number of (nominally) uniformly distributed baselines. In the S-band case, single-pass Pol-InSAR acquisitions are available as well, allowing to assess directly temporal decorrelation effects. Field inventory data collected continuously in space over 25 ha at single-tree level and fine-beam airborne lidar data are used as a reference. References [1] J. Du, J. Shi, R. Sun, “The Development of HJ SAR Soil Moisture Retrieval Algorithm,” Internationa Journal of Remote Sens., vol. 31, 2010, pp. 3691–3705. [2] R. Bird, P. Whittaker, B. Stern, N. Angli, M. Cohen, R. Guida, “NovaSAR-S: A Low Cost Approach to SAR Applications,” Proc. APSAR 2013, Tsukuba, Japan, Sept. 2013. [3] P. Rosen, Y. Kim, R Kumar, T. Misra, R. Bhan, V. Raju Sagi, “Global Persistent SAR Sampling With the NASA-ISRO SAR (NISAR) Mission,” Proc. of IEEE RadarConf 2017, Seattle, WA, USA, May 2017. [4] A. Natale, R. Bird, P. Whittaker, R. Guida, M. Cohen, D. Hall, “Demonstration and Analysis of the Application of S-Band SAR,” Proc. IGARSS 2011, Vancouver, Canada, Jul. 2011. [5] M. Pinheiro, A. Reigber, R. Scheiber, P. Prats-Iraola, A. Moreira, “Generation of Highly Accurate DEMs Over Flat Areas by Means of Dual-Frequency and Dual-Baseline Airborne SAR Interferometry,” IEEE Trans. Geosci. Remote Sens., 56(8), Aug. 2018, pp. 4361-4390. [6] R. K. Ningthoujam et al., “Airborne S-Band SAR for Forest Biophysical Retrieval in Temperate Mixed Forests of the UK,” Remote Sensing, 2016, 8(7), 609; doi:10.3390/rs8070609 [7] M. Pardini, K. Papathanassiou, “First Investigation on the Information Content of Multibaseline Pol-InSAR Data at S-Band for Forest Structure Observation,” Proc. IGARSS 2015, Milan, Italy, Jul. 2015.
15:00 -
Combined use of PolSAR and PolInSAR based indicators dedicated to forest biomass estimation from P-Band multibaseline SAR data
GELAS, Colette (1); VILLARD, Ludovic (1); KOLECK, Thierry (1); LE TOAN, Thuy (1); DANIEL, Sandrine (2); POLIDORI, Laurent (1) - 1: CESBIO, France; 2: Capgemini, France
The BIOMASS mission has been designed to provide forest above ground biomass (AGB) and forest height maps at continental scales using PolSAR, PolInSAR and TomoSAR acquisitions. One of the challenging topic for the retrieval methods under development is to take the most of these data and its combined use, considering the TomoSAR stack and the triplets of PolInSAR data acquired respectively during the so-called TOM and INT phases of the mission. This study proposes the implementation of a method combining two indicators referred to as t0 and hC, respectively based on PolSAR and PolInSAR data. The t0 indicator has been developed in order to minimize the perturbing effects caused by terrain topography using a twofold correction based on the polarimetric information contained on the coherency matrix and the knowledge of local geometry derived from the digital elevation model (DEM). The indicator hC is the PolInSAR height resulting from the retrieval method based on the RVoG model, accounting for geometrical corrections and enhanced by the use of ground height and extinction coefficients from TomoSAR data. Relations between these indicators and forest AGB are constructed with likelihood models based on adaptive spreads around poly-logarithmic functions, in order to reproduce the join probability function between t0 and hC. While the latter poly-log functions are derived from training plots, the adaptive spread is generated from the MIPERS4D model, which is parametrized using ecological database and earth observation data such as DEM and precipitation data in order to infer the local ground topography and vegetation water content and soil moisture. Initiated by the BRIX initiative, the proposed method has been implemented on the so-called Test Bed, a remote virtual machine implemented by ESA that provides a homogeneous access to all BIOMASS campaign data and allows to implement and run algorithms in common open computing languages. Study cases derived from TropiSAR and AfriSAR airborne campaigns have been used in order to test the algorithm transferability, with a particular focus on the chosen training plots which are needed for the poly-log models. In addition to significant improvements related to the correction of perturbing effects caused by terrain topography, the combined use of t0 and hC provides more accurate results for the high values of AGB, and more stable results across the test sites including dense forests, with however similar or decreased performances in some cases of lower AGB based on a separate use of t0 or hC. Although achieved on tropical test sites and assuming minor temporal effects, this method is also applicable to other types of forest, and perturbing effects due to temporal changes are expected to be significantly decreased given previous work on the interdependency between the variations of backscattering coefficients and temporal decorrelation.
15:20 - 15:40
15:40 - 17:20
15:40 -
The concept of multi-Mission Algorithm and Analysis Platform (MAAP) for the BIOMASS mission
Albinet, Clément (1); Frommknecht, Björn (1); Laur, Henri (1); Costa, Gabriella (1); Whitehurst, Amanda (2); Murphy, Kevin (2); Scipal, Klaus (3) - 1: ESA-ESRIN, Frascati, Italy; 2: NASA-HQ, Washington, USA; 3: ESA-ESTEC, Noordwijk, The Netherlands
With the launch of new satellite missions and growing understanding of the complexity of ecological processes, the scientific community is faced with a unique and immediate need for improved data sharing and collaboration. This is especially evident in the Earth sciences and carbon monitoring community with the launch of the ESA BIOMASS mission [1], the NASA-ISRO SAR (NISAR) mission [2], and the NASA Global Ecosystem Dynamics Investigation (GEDI) mission [3]. While these missions and the corresponding research leading up to launch, which includes airborne, field, and calibration/validation data collection and analyses, provide a wealth of data and information relating to global biomass estimation, they also present data storing, processing and sharing challenges. The NISAR mission alone will produce around 40 petabytes of data per year. Due to the constraints of existing organizational infrastructures, these large data volumes will place accessibility limits on the scientific community and may ultimately impede scientific progress. In this context, the concept of ESA-NASA multi-Mission Algorithm and Analysis Platform (MAAP) dedicated to the BIOMASS, NISAR and GEDI missions is proposed. This analysis platform will be a virtual open and collaborative environment. The goal is to bring together data centre (Earth Observation and non- Earth Observation data), computing resources and hosted processing, collaborative tools (processing tools, data mining tools, user tools, …), concurrent design and test bench functions, application shops and market place functionalities, accounting tools to manage resource utilisation, communication tools (social network) and documentation. The goal for the MAAP is to establish a collaboration framework between ESA and NASA to share data, science algorithms and compute resources in order to foster and accelerate scientific research conducted by NASA and ESA EO data users. The objectives of the MAAP for BIOMASS, NISAR and GEDI missions are to: 1) Enable researchers to easily discover, process, visualize and analyze large volumes of data from both agencies; 2) Provide a wide variety of data in the same coordinate reference frame to enable comparison, analysis, data evaluation, and data generation; 3) Provide a version-controlled science algorithm development environment that supports tools, co-located data and processing resources; 4) Address intellectual property and sharing issues related to collaborative algorithm development and sharing of data and algorithms. REFERENCES [1] T. Le Toan, S. Quegan, M. Davidson, H. Balzter, P. Paillou, K. Papathanassiou, S. Plummer, F. Rocca, S. Saatchi, H. Shugart and L. Ulander, “The BIOMASS Mission: Mapping global forest biomass to better understand the terrestrial carbon cycle”, Remote Sensing of Environment, Vol. 115, No. 11, pp. 2850-2860, June 2011. [2] P.A. Rosen, S. Hensley, S. Shaffer, L. Veilleux, M. Chakraborty, T. Misra, R. Bhan, V. Raju Sagi and R. Satish, "The NASA-ISRO SAR mission - An international space partnership for science and societal benefit", IEEE Radar Conference (RadarCon), pp. 1610-1613, 10-15 May 2015. [3] https://science.nasa.gov/missions/gedi
16:00 -
BRIX: The first Biomass Retrieval Inter-comparison eXercise
Albinet, Clément (1); Balzter, Heiko (2); da Conceição Bispo, Polyanna (2); García, Mariano (3); Gelas, Colette (4); Notarnicola, Claudia (5); Pacheco-Pascagaza, Ana María (2); Padovano, Antonio (5); Paloscia, Simonetta (6); Pettinato, Simone (6); Pourshamsi, Maryam (2); Rodríguez-Veiga, Pedro (2); Santi, Emanuele (6); Scipal, Klaus (7); Villard, Ludovic (4) - 1: ESA-ESRIN, Frascati, Italia; 2: University of Leicester, Leicester, United Kingdom; 3: University of Alcalá, Alcalá de Henares, España; 4: CESBIO, Toulouse, France; 5: EURAC, Bolzano, Italia; 6: IFAC-CNR, Firenze, Italia; 7: ESA-ESTEC, Noordwijk, The Netherlands
Biomass Retrieval Inter-comparison eXercise (BRIX) is a biomass retrieval algorithm inter-comparison exercise. It was conducted by European Space Agency (ESA) and multiple participants to develop algorithms for biomass estimation. The exercise aimed at using Synthetic Aperture Radar (SAR) P-Band datasets acquired as part of the ESA’s SAR campaigns in support of the upcoming ESA’s BIOMASS mission [1]. ESA set up the BIOMASS Algorithm Test Bed which is a virtual machine that (I) provides access to all BIOMASS campaign data in a unified format; (II) includes software tools that allow to implement and run algorithms in common open programming languages (Python, C, Fortran) as well as a limited number of proprietary programming languages (IDL, Matlab); and (III) makes available processing resources. Participants were invited to upload or develop their code, and run it on the Test Bed using the predefined campaign datasets. BRIX consisted of two phases. In the first phase, four different retrieval algorithms were developed based on the data acquired during the AfriSAR campaign. One of the recent ESA’s SAR campaign that was conducted in Gabon in 2016. The SAR P band data acquired over four test sites: Lopé (super site), Mondah, Rabi and Mabounie. A complete set of data was made available to participants for Lopé including Tomographic SAR Stacks, Polarimetric SAR, Polarimetric Interferometric SAR, Digital Terrain Model (DTM), which were used as independent datasets in the models, and LiDAR derived Biomass, Canopy Height Model (CHM) and field measured data, used as reference data. In the second phase, the algorithms were assessed, on three other sites (Mondah, Rabi and Mabounie). There, only a limited set of data was available and no reference data was provided. ESA compared and evaluated the derived biomass maps over the three test sites with existing ground data and LiDAR derived biomass. The tested approaches include physically based models, empirical and machine learning techniques. During the BRIX exercise following challenges were raised: 1. How to optimally use the available Polarimetric, Polarimetric Interferometric and Tomographic SAR data in one consistent retrieval concept. 2. The transferability of the retrieval developed over one site (Lopé) to three sites that are separated by several 100 km, which requires to account for different ecological characteristics and different forest structure. 3. How to correct the topographic effects inside SAR images. REFERENCES [1] T. Le Toan, S. Quegan, M. Davidson, H. Balzter, P. Paillou, K. Papathanassiou, S. Plummer, F. Rocca, S. Saatchi, H. Shugart and L. Ulander, “The BIOMASS Mission: Mapping global forest biomass to better understand the terrestrial carbon cycle”, Remote Sensing of Environment, Vol. 115, No. 11, pp. 2850-2860, June 2011.
16:20 -
MACHINE LEARNING APPLICATION FOR THE FOREST BIOMASS RETRIEVAL IN THE FRAMEWORK OF ESA BRIX
Santi, Emanuele (1); Paloscia, Simonetta (1); Pettinato, Simone (1); Cuozzo, Giovanni (2); Padovano, Antonio (2); Notarnicola, Claudia (2) - 1: IFAC - CNR, Italy; 2: EURAC, Italy
In this study, two different machine learning approaches aimed at estimating the forest biomass (t/ha) from the ESA airborne SAR missions, namely Artificial Neural Network (ANN) [1-2], and supported Vector Regressions (SVR) [3-4], have been implemented and validated. The outputs of these two approaches have been compared with random forests (RF), an ensemble learning method that is very popular for classification, and retrieval applications. This activity has been carried out in the framework of the BRIX exercise, with the aim of intercomparing biomass retrieval algorithms for P-band full-polarimetric SAR sensors in view of the upcoming ESA BIOMASS mission (a P-band synthetic aperture polarimetric radar). Several strategies have been exploited, by developing “general” algorithms trained with data derived from the whole dataset and “specific” algorithms, trained with data derived from a single campaign, among Afrisar, Biosar and Tropisar. In all cases, the algorithms have been trained on a subset of the available data and validated on the remaining, obtaining correlation coefficients between R=0.82 and R= 0.94, with a RMSE between 15 t/ha and 70 t/ha, depending on the algorithm and on the dataset. In the case of ANN, the validation of the “general” and “specific” algorithms resulted in a correlation coefficient between R=0.78 and R=0.94, depending on the dataset, with a RMSE between 15 and 60 t/ha and negligible BIAS. The validation of the SVR algorithms resulted in a correlation coefficient between R=0.27 and R=0.90, depending on the dataset, with a corresponding RMSE between 25 and 77 t/ha and BIAS negligible in this case too. After validation, both ANN and SVR algorithms have been applied to the entire SAR images for generating the corresponding biomass maps. The comparison of ANN, SVR and RF methods helped in pointing out the advantages and disadvantages of each technique in terms of retrieval accuracy, exportability to other datasets and computational cost. References [1]. Santi E., S. Paloscia, S. Pettinato and G. Fontanelli. “Application of artificial neural networks for the soil moisture retrieval from active and passive microwave spaceborne sensors,” Int. J. Appl. Earth Observ. Geoinf., vol 48, pp. 61–73, Jun. 2016. [2]. Santi E., S. Paloscia, S. Pettinato, G. Fontanelli, M. Mura, C. Zolli, F. Maselli, M. Chiesi, L. Bottai, G. Chirici, 2017, The potential of multifrequency SAR images for estimating forest biomass in Mediterranean areas, Remote Sensing of Environment 200 (2017), pp. 63–73. [3]. Pasolli, L., Notarnicola, C., Bruzzone, L., Bertoldi, G., Della Chiesa, S., Hell, V., Niedrist, G., Tappeiner, U., Zebisch, M., Del Frate, F., Vaglio Laurin, G. 2011. Estimation of Soil Moisture in an Alpine Catchment with RADARSAT2 Images. Hindawi Publishing Corporation, Applied and Environmental Soil Science, Article ID 175473, 12 pages, doi:10.1155/2011/175473 [4]. Pasolli, L., Notarnicola, C., Bertoldi, G., Bruzzone, L., Remelgado R., Greifeneder, F., Niedrist, G., Della Chiesa, Tappeiner, U., Zebisch, M. 2015. Estimation of Soil Moisture in Mountain Areas Using SVR Technique Applied to Multiscale Active Radar Images at C-Band. IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING, VOL. 8, NO. 1, JANUARY 2015, doi: 10.1109/JSTARS.2014.2378795.
16:40 -
Multibaseline PolinSAR in Mangrove Forests of Gabon
simard, Marc (1); Denbina, Michael (1); Fatoyinbo, Lola (2); Pinto, Naiara (1) - 1: Jet Propulsion Laboratory, California Institute of Technology, United States of America; 2: Goddard Space Flight Center, USA
The AfriSAR campaign is a joint NASA and European Space Agency airborne campaign conducted in Gabon in support of the upcoming ESA BIOMASS, NASA-ISRO Synthetic Aperture Radar (NISAR) and NASA Global Ecosystem Dynamics Initiative (GEDI) missions. In this paper, we present results based on data collected by the UAVSAR airborne L-band system acquired in repeat-pass interferometric mode and Laser Vegetation and Ice Sensor (LVIS) airborne Lidar. The UAVSAR data is available from https://uavsar.jpl.nasa.gov, while the LVIS data is available from https://lvis.gsfc.nasa.gov/. We introduce a new methodology based on optimizing the baseline selection through a machine learning algorithm trained on the Lidar estimates of canopy height. The proposed baseline selection methods use various data quality metrics calculated from the PolInSAR coherences and other parameters. We pose the problem, for simplicity, as a classification exercise solved with machine learning. The implementation uses a Support Vector Machine (SVM) library to classify polinSAR metrics for each pixel into the best interferometric baseline. The forest canopy height is then estimation with this best polinSAR baseline. Training of the SVM was performed with the Lidar dataset and validated with in situ data. We present results obtained over mangrove sites of the Pongara and Akanda National Parks. These forests are characterized by a wide range of heights and structure clearly impacting the relative contribution of scattering mechanisms. Overall, we found errors on the order of 4m or about 10%.
17:00 -
Mangroves detection and classification using full-polarimetric SAR data
Nunziata, Ferdinando (1); Ferrentino, Emanuele (1); Migliaccio, Maurizio (1); Zhang, Hongsheng (2) - 1: Università di Napoli Parthenope, Italy; 2: The Chinese University of Hong Kong, China
Mangrove forests are coastal wetlands that contribute to biodiversity and act as major biogeochemical links between upland and coastal regions. For this reason, they play a major part in the costal ecosystem and sea coast conservation. Classification of mangrove species can be very important in formulating an inventory for use in the development of conservation management plans. However, the classification, typically involves intensive field surveys, and this can be very expensive and hard to undertaken. Within this context, microwave remote sensing, and in particular the Synthetic Aperture Radar, can be a very useful and cheap tool for mangroves observation, since its guarantees all-day and almost all-weather synoptic observations of the Earth’s surface. The main goals of this study are to develop polarimetric methods to classify scattering mechanisms that characterize mangroves using full- and dual-polarimetric SAR data. For this purpose, methods based on polarimetric model decompositions (i.e. Cloude Pottier and Freeman-Durden decomposition) are used in order to detect and classify different mangrove species together with the Pauli phase which, carrying on information on the electric properties of the observed scene, is expected to discriminate among the different mangrove types. Preliminary results are obtained processing a set of SAR data collected over the Mai Po Marshes Nature Reserve, located in Hong Kong, by the C-band Alos-2 mission show that the proposed methodologies, well detect and classify the different mangroves species.
17:20 - 17:40
09:30 - 10:30
09:30 -
Precise Detection of Oil Palm Trees Areas by a single Fully Polarimetric Synthetic Aperture Radar Image and Multi Chromatic Analysis
Biondi, Filippo - University of L'Aquila, Italy
The objective of this research is to perform precise detection of oil palm trees areas using a single fully polarimetric synthetic aperture radar (SAR) image. The expansion of palm oil plantations across the earth is causing deforestation of natural rain forest and conversion of peat land into plantation land. This research aims to give reliable contribute in contrasting the ongoing increase in carbon dioxide (CO2) emissions over the earth atmosphere. This paper presents an improvement of the PolSAR decomposition scheme which permits the performing of more accurate classification. The method which use additive information existing by the interference generated between two Doppler sub-apertures SAR images processed by one SAR observation. This interferometric polarimetric SAR (PolInSAR) multi-chromatic analysis (MCA-PolInSAR) signal processing method permits the efficient separation of different types of trees. In this case the oil palms have lower height and density because of their different structure respect to the other trees populating the natural forest areas. Results also show a considerable improvement in robustness of classification, in terms of definition and precision.
09:50 -
DOUBLE-BOUNCE CONTRIBUTION EFFECT IN THE ESTIMATION OF BIOPHYSICAL PARAMETERS OF VEGETATION BASED ON POLINSAR TANDEM-X BISTATIC DATA
Romero-Puig, Noelia; López-Sánchez, Juan M.; Ballester-Berman, J. David - Instituto Universitario de Investigación Informática, Universidad de Alicante, Spain
The inversion of the well-known Random Volume over Ground (RVoG) model [Treuhaft1996, Treuhaft2000, Papathanassiou2001] is employed for the estimation of physical parameters of scenes with vegetation by exploiting Polarimetric SAR Interferometry (PolInSAR) data [Cloude1998]. Data gathered by the TanDEM-X satellite formation [Krieger2007] are characterised by a single-pass bistatic configuration, where one satellite is transmitting and both of them are receiving, i.e. there is one monostatic image and one bistatic image. As a result from this bistatic configuration, the formulation of the interferometric coherence accounts for an extra decorrelation term: a double-bounce contribution at the ground which entails also volume effects from the interferometric point of view [Treuhaft1996]. This double-bounce decorrelation factor has been overlooked in previous works exploiting TanDEM-X data on vegetation height estimation in forests [Kugler2014, Kugler2015, Lee2015, Abdullahi2016], and only considered in the inversion of the RVoG model over rice fields [Lopez2017]. In this work we provide a detailed analysis of the effect of the double-bounce decorrelation factor on the inversion of scene parameters, with particular focus on the vegetation height. The study employs both simulated data as well as real data acquired over rice fields during the science phase of the TanDEM-X mission. The potential limitations of current inversion approaches are assessed, and the influence of both system parameters (i.e. incidence angle) and scene parameters (i.e. extinction coefficient and ground-to-volume ratios) is evaluated. Results show that the bias in the estimation of scene parameters is higher when the incidence angle is above 30 degrees, i.e for shallow incidences. The normalised vegetation height, i.e. expressed as $k_v$, is used in order to extrapolate the results to other scenarios, e.g. forests.
10:10 -
POL-timeSAR: Benefits of polarimetry on change detection in time-series
KOENIGUER, Elise - Onera, France
With the development of open-source data, it is becoming increasingly easy to obtain a large number of images on a single site, especially thanks to the Sentinel 1 satellites of the Copernicus mission. This change of context makes it possible to envisage a better temporal monitoring, whether for environmental needs (monitoring of forests, glaciers), industrial (crop management, urban planning), or surveillance (maritime traffic, camp establishment). Recently, a method for visualizing changes occurring during a particular period of time (Reactiv, [1],[2]) has been developed at Onera. It is based on the use of SAR time series and has been created after the analysis and modeling of temporal statistical properties of speckle and permanent scatterers. This method has been already successfully tested on Sentinel 1 and TerraSAR-X images, both on locally downloaded data and applied to the Google Earth Engine platform. It can be also transformed into an automatic detection of areas subject to particular evolutions. However, up to now, only single polarimetric time-series have been considered. Therefore, the object of this article is to question the influence of polarimetry, on the visualization and on the detection of temporal changes in large SAR time-series. Legitimate questions about the use of polarimetry are numerous: - how does the polarization affect the result of the colored composition obtained? - which polarization is most effective to robustly detect changes or temporal activity? - do the different polarimetric channels provide complementary information, or are they largely correlated? - can polarimetry be used to help the interpretation of detected changes? - must it be applied before or after change detection? To answer these questions, two types of data have been used in this study: full polarimetric data on a stack of San Francisco restricted to 12 images; but also partially polarized Sentinel 1 data at various sites using GRD images on the Google Earth platform, and SLC data downloaded on Saclay (France) for which an extensive ground truth has been made. First, the visualization algorithm is outlined. Results obtained on the same site for the different polarizations are compared for long events such as construction sites. The results show that all polarimetric channels highlight the same changes. Then, we compare the influence of the polarization on the temporal profiles obtained for natural soils subjected to variations such as agricultural fields. They show the complementarity of the different polarimetric channels. A new algorithm that includes the different available polarimetric signals is proposed. Finally, polarimetry is also studied for "point-events" detection, that means some high signal occurring only once during the observed duration, for example, a vehicle. We show that most of the time, only one polarization is able to find such an event, and thus that full polarimetry is required to detect all these point-events. [1] Koeniguer, E. C., Nicolas, J. M., Pinel-Puyssegur, B., Lagrange, J. M., & Janez, F. Visualisation des changements sur séries temporelles radar: méthode REACTIV évaluée à l’échelle mondiale sous Google Earth Engine, Revue Française de Photogrammétrie et télédétection, 2018 (Best Paper Award for the RFIAP-CFPT conference (2018)) [2] Colin-Koeniguer, E., Boulch, A., Trouve-Peloux, P., & Janez, F. (2018, June). Colored visualization of multitemporal SAR data for change detection: issues and methods. In EUSAR 2018; 12th European Conference on Synthetic Aperture Radar (pp. 1-4). VDE.
10:30 - 11:00
11:00 - 12:00
11:00 -
Nonlocal filtering of polarimetric SAR images applied to change detection for volcano monitoring
D'Hondt, Olivier; Valade, Sébastien; Hellwich, Olaf - TU Berlin, Germany
This work presents the application of a new multi-channel SAR nonlocal filtering framework to incoherent change detection for volcano monitoring. Synthetic Aperture Radar is particularly useful for volcano monitoring, as it provides measurements when no visual observations are possible, i.e. at night and in poor weather conditions. Recent satellite constellations such as Sentinel-1 provide unprecedented temporal samplings, boosting the potential for operational monitoring with this type of data. In particular, change detection techniques applied to SAR imagery can be used to assess volcanic hazards and map eruptive features (i.e., detect dome growth, map flow extents, etc.). In regions where strong decorrelation are expected (due to dense vegetation in tropical regions, or ice/snow cap at high altitudes/latitudes), coherence-based change detection approaches will fail, and require using amplitude-based (incoherent) approaches. Nevertheless, delineating precise boundaries of changed regions remains challenging, in particular because of the intrinsic speckle noise and the complexity of SAR scattering responses. These difficulties are here tackled using a nonlocal filtering approach exploiting the dual-polarimetric information. Due to the presence of speckle, incoherent change detection of distributed targets has to be considered in a statistical framework. The computation of statistical change detectors requires an averaging over several samples. On real data this average is perform in a spatial neighborhood of the pixel to consider. Spatial averaging by boxcar filtering is only valid in areas where the statistical properties of neighbor pixels are spatially stationary. In practice, real data exhibits a nonstationary spatial structure due to the presence of edges between homogeneous regions. Therefore, spatially adaptive filtering is required to preserve the local structure of the scene. At volcanoes, lava and pyroclastic flows may concentrate in narrow valleys, thereby defining elongated structures which are blurred by traditional boxcar filtering. In this work we propose to apply a recently developed filter based on the nonlocal principle to circumvent this problem. This method allows a spatially adaptive estimation of the polarimetric covariance matrix thanks to the use of matrix-based similarities and exploits the texture information present in the image intensity. The filtering procedure is composed of two stages: A first stage allows a pre-estimation of the covariance matrix with a fixed number of looks thanks to a region growing procedure. This step preserves the spatial resolution as opposed to the common pre-summing step required on single-look complex data. Then, a second step applies the nonlocal principle to the pre-filtered covariance leading to the final covariance estimate. Both pre- and post-event images are filtered in a single step by considering the multi-temporal polarimetric covariance. It is then possible to apply polarimetric change detectors on the individual de-speckled images. We demonstrate the effects of the filtering method on change detection with dual-pol Sentinel-1 data for different image pairs acquired at different erupting volcanoes. We compare the output of change detection when our method has been applied with results obtained with boxcar filtering. Our approach allows a better contrast between change and no-change areas than boxcar filtering and succeeds in preserving the anisotropic structure of thin regions related to lava or pyroclastic flows emplaced in vegetated areas.
11:20 -
Monitoring and detecting archaeological features with multi-frequency polarimetric analysis
Patruno, Jolanda (1); Delgado Blasco, Jose Manuel (2); Fitrzyk, Magdalena (3) - 1: Rhea Group c/o ESA/ESRIN, Via Galileo Galilei 00044 Frascati, Italy; 2: Universidad de Jaén, Campus Las Lagunillas s/n, 23071 Jaén, Spain; 3: RSAC c/o ESA/ESRIN, Via Galileo Galilei 00044 Frascati, Italy
Nowadays, SAR applications for archaeology focus on high spatial resolution SAR sensors, which allow the recognition of structures of small dimension and give information of the surface topography of sites. Given the potential of combined dual and fully polarised SAR data, the analysis focuses on a polarimetric multi-frequency and multi-incidence angle analysis of Sentinel-1 (C-band), ALOS PALSAR (L-band) and of RADARSAT-2 (C-band) sensors for the detection of surface and subsurface archaeological structures over the UNESCO site of Gebel Barkal (Sudan). While PALSAR offers a good historical reference, Sentinel-1 time series provide recent and systematic monitoring opportunities. RADARSAT-2 polarimetric data have been specifically acquired in 2012/2013, and have been scheduled to achieve a multi-temporal observation of the archaeological area. Sensors spatial resolution (Sentinel-1, ALOS PALSAR ca. 20 m and RADARSAT-2 ca. 10 m), archaeological structures’ dimensions, morphology and environment and the great potential L-band and C-band demonstrated in such context, make the research suitable for the monitoring of the archaeological area, as well of the possible threatening factors that can affect the integrity of a cultural site. In remote sensing for archaeology, an unequivocal method capable of an automatic detection of archaeological features is still not existing. Hence, objective of the work is to exploit the potential of such complex but meaningful technique as SAR polarimetry is, and to individuate investigation guidelines thanks to the combined use of dual and full pol dataset for a possible use in the archaeological domain.
11:40 -
Cultural effects on the Urban detection based on SAR polarimetric characteristics
Blasco, Jose Manuel (1); Patruno, Jolanda (2); Fitrzyk, Magdalena (3) - 1: Grupo de Investigación Microgeodesia Jaén (PAIDI RNM-282), Universidad de Jaén, Campus Las Lagunillas s/n, 23071 Jaén, Spain; 2: Rhea Group c/o ESA/ESRIN; 3: RSAC c/o ESA/ESRIN
Urban planning and cultural behaviour varies depending on the part of the world you look at. In this work, we want to show how different are cities from the Polarimetric SAR point of view, and for that we have compared European with American and African cities using L and C frequency bands. Specific datasets are full and dual polarisation of ALOS PALSAR and the dual pol Sentinel-1 over San Francisco (USA), Milan (Italy) and Cairo (Egypt). Using the full-polarised ALOS PALSAR data we have observed that on the case of San Francisco, urban and non-urban can be easily differentiated, as the double-bounce polarimetric signal prevails on urban environment, being not the same case for cities as Cairo, where the signal characteristics from the city does not represent only the double-bounce, but also other kind of scattering. We have investigated these phenomena over Cairo using the cross-polarised channels VH of both Sentinel-1 and ALOS PALSAR data, to understand whether VH signal could be affected by oriented targets which could create also high cross-pol channels. Traditional classification methods based only on the different polarimetric signatures seems not to be enough to properly classify the different urban environments on which urban signatures does not respond to the expected as for European and American cities
12:00 - 12:20
12:20 - 13:40
13:40 - 15:20
13:40 -
Investigating Detection Capabilities of Distressed Refugee Boats Using SAR
Lanz, Peter (1); Marino, Armando (2); Brinkhoff, Thomas (3); Köster, Frank (4); Möller, Matthias (5) - 1: Dept. of of Computing Science, Carl von Ossietzky Univ. of Oldenburg / Institute for Applied Photogrammetry and Geoinformatics, Jade Univ. Oldenburg; 2: Department of Biological and Environmental Sciences, University of Stirling; 3: Institute for Applied Photogrammetry and Geoinformatics, Jade University of Applied Seiences; 4: Institute of Transportation Systems, German Aerospace Center (DLR); 5: Faculty for Humanities and Cultural Sciences, Otto-Friedrich-University of Bamberg
PLEASE SEE ATTACHED FILE! Existing research in (semi-)automatic marine target detection using Synthetic Aperture Radar (SAR) data mainly concentrates on the detection and classification of large, metallic targets - mainly ships. This work focuses on the detection of small, non-metallic targets, in particular inflatable rubber vessels. Such vessels are used by migrants attempting to cross from Africa to Europe. The physical attributes of such kind of targets, namely its small size and height and the absence of materials of high dielectric constant such as metals, decrease the detection capabilities of commonly known vessel detection systems. In this work we applied and tested a range of detectors and different methodology to gain a better understand of the target’s backscattering properties. The goal is to identify the best way to proceed in the effort of developing a new, specially tailored detection algorithm. We use multi-platform SAR data (mainly TerraSAR-X, accompanied by Senitel1 & Sentinel2) holding “sea truth” to a 12m long rubber inflatable vessel which was used in 2015 by migrant to cross the Mediterranean. The data were collected in the Müggelsee, a lake near Berlin, Germany, which functioned as a test bed. A lake scenario was selected because it allows clear identification of the scattering mechanisms of inflatable boats (since the background is generally lower than for open ocean). The collection comprises single and dual-polarimetric data and covers a variety of different sensor and geometry parameters. Different combinations of incidence angles, acquisition modes, targets orientation relative to the LoS, movement and cargo supports the effort of identifying the main scattering mechanism and chances and limitation of its detectability. In this preliminary work we applied, apart from the intensity-based CA-CFAR detector (Fig. 2), Polarimetry-based and sub-look-based detectors following a qualitative evaluation of their detection capabilities regarding the special target. Amongst the Polarimetry detectors are the intensity Depolarization Ratio Anomaly Detector (iDPolRAD) (Marino et al. 2016), the Geometrical Perturbation-Polarimetric Notch Filter (GP-PNF) (Marino, 2013), the Polarimetric Match Filter (PMF), the polarimetric symmetry and entropy detectors. The sub-look detectors include the sub-look coherency (Marino et. al. 2015), the sub-look entropy (Schneider et.al. 2006) and the sub-look product. As additional step, the iDPolRAD was modified in order to focus the detector on surface scattering anomalies instead than volume anomalies. This is useful when the water has a low backscattering and the target is mostly represented by surface scattering. We finally considered a combination of the two iDPolRAD to more efficiently detect the inflatables. The novelty of this work lies in two points: a) it is the first time that a broad variety of vessel detectors are used to detect and analyse this kind of special inflatables target; b) the iDPolRAD was modified to detect anomalies in surface scattering. The well-established CA-CFAR is capable of detecting the target (Figure 2) in several cases but can be biased by strong sea surface clutter and reduced target radar response. Main factors reducing the target’s signal can be (a combination of) a low incidence angle, the cross polarimetric channel and the vessel’s orientation. The adapted version of iDPolRAD, using the combination of the two scattering mechanisms volume and surface scattering, increases detection capabilities for very small targets. Fig. 3 shows first results. To reduce the rate of false positives of such a detector, a preliminary analysis of the sea surface scatter has to be done. In case of low clutter, the main signal is produced by thermal noise. This state of high Entropy allows for target identification when there are clearly identifiable surface scattering mechanisms. In case of high surface scattering levels, the detector only searches for volume scattering. The notch filter, employing a geometrical perturbation analysis, shows clear differences in the polarimetric signature between water and the target (Fig. 4). The entropy detector on the other hand left us with rather unsatisfyingly results because the water had low backscattering and the thermal noise produced very high entropy. The polarimetric symmetry detector reveilles the target’s asymmetric behaviour (Fig. 5). The PMF shows the best results in terms of contrast (Fig. 6). Very promising results can be seen from the sub-look product algorithm, both in terms of contrast and size (Fig 7). A quantitative analysis of the different detectors’ capabilities is still under progress (and it will be ready for the conference) and this abstract only includes first results. The primary motivation of this research is to mitigate the ongoing humanitarian crisis at Europe's southern Sea border. The applicability of the project’s results to the setting of the open sea, where stronger winds and seas could interfere with radar detection, will be discussed. This project builds a foundation to develop satellite based detection systems for inflatable rubber boats. Such systems could be integral to search and rescue infrastructure in reducing the number of lives lost at sea. References: Marino, A. (2013). A notch filter for ship detection with polarimetric SAR data. Selected Topics in Applied Earth Observations and Remote Sensing, IEEE Journal Of, 6(3), 1219–1232. Marino, A., Sanjuan-Ferrer, M. J., Hajnsek, I., & Ouchi, K. (2015). Ship Detection with Spectral Analysis of Synthetic Aperture Radar: A Comparison of New and Well-Known Algorithms. Remote Sensing, 7(5), 5416–5439. https://doi.org/10.3390/rs70505416 Schneider, R. Z., Papathanassiou, K. P., Hajnsek, I., & Moreira, A. (2006). Polarimetric and interferometric characterization of coherent scatterers in urban areas. IEEE Transactions on Geoscience and Remote Sensing, 44(4), 971–984. https://doi.org/10.1109/TGRS.2005.860950 Marino, A., & Hajnsek, I. (2014). A Change Detector Based on an Optimization With Polarimetric SAR Imagery. IEEE Transactions on Geoscience and Remote Sensing, 52(8), 4781–4798. https://doi.org/10.1109/TGRS.2013.2284510 Marino, A., Dierking, W., & Wesche, C. (2016). A Depolarization Ratio Anomaly Detector to Identify Icebergs in Sea Ice Using Dual-Polarization SAR Images. IEEE Transactions on Geoscience and Remote Sensing, 54(9), 5602–5615. https://doi.org/10.1109/TGRS.2016.2569450
14:00 -
Retrieval of the dielectric properties of oil slick using SAR via a Polarimetric Two-Scale Model
Quigley, Cornelius Patrick; Eltoft, Trobjørn - UiT The Arctic University of Norway, Norway
Marine oil spills represent an important environmental problem. Surveillance systems such as SAR that can accurately characterise key features of oil slicks, such as volumetric content, is of vital importance. This study focuses on investigating the capability of a Polarimetric Two-Scale Model (PTSM) to retrieve an estimate for the dielectric constant of oil slicks. In this model the ocean surface is considered to be composed of a collection of slightly rough, randomly titled facets. The scattering from each facet is assumed to be in accordance with the Small Perturbation Model (SPM). By averaging over the slopes of the facets in the range and azimuth direction, expressions for the Normalised Radar Cross Section (NRCS) are derived that are more easily able to describe the scattering from complex surfaces such as the ocean-atmosphere interface. From previous studies, it has been shown that this model provides more accurate results in the estimation of the dielectric constant for soil moisture retrieval than the simple SPM or the Extended Bragg model (XBragg). This model has also proven itself to be able to account for both cross-polarisation and depolarisation effects while also maintaining a simple formulation. While this model has previously been applied to the estimation of soil moisture content, it has not been applied to the estimation of the dielectric properties of oil slick. A collection of Radarsat-2 quad-polarimetric data sets are used that were acquired off the Norwegian coast by the Norwegian Clean Seas Association for Operating Companies (NOFO) when they conducted their annual oil-on-water exercise in the North Sea. Data sets were acquired during the summer months of 2011, 2012, 2013, 2015 and 2016. As part of the exercises, variable substances were included into the marine environment and imaged side-by-side. These include crude oil, emulsions and plant oil simulators. Given the difficulty in obtaining SAR imagery of verified mineral oil/plant oil slicks, these data sets represent a unique opportunity in investigating the behaviour of oil slick when imaged from SAR sensors. The data taken by the sensor were used in conjunction with the PTSM in order to invert for an estimate for the dielectric constant. Radar returns from slick are typically less than the surrounding ocean due to the reduction in surface roughness and can be highly influenced by sensor noise. Due to the heavy effect noise can have on the estimation, the model is adopted from its original form to make use of only the co-polarization channels. The cross-polarisation channels are used to estimate the thermal noise present and subtract it from the data before an inversion is performed. When applying the model the co-polarization ratio, i.e. VV/HH, is employed in order to remove the dependancy on the small-scale roughness. By applying the model in this way the only factors that are pertinent to the model are the epsilon value, the large-scale roughness parameter and the Hurst coefficient, a parameter related to the fractal dimension of the scattering surface. Both epsilon and the large-scale roughness descriptor are estimated by building up numerical charts that fit the data well which can then be read off. The Hurst coefficient has been shown to have a weak effect on the model and so is estimated heuristically. The results of the the inversion show that there is a high contrast between the slick and the surrounding ocean up until an epsilon value of about 20. Within the slick, estimations for the absolute value of epsilon range from 1 to 20 with the slick showing internal zoning, indicating that areas where there is higher volumetric portions oil can be determined. Those scenes that contain both plant oil simulators and crude oil show differences in the proportion of lower epsilon estimates. This is likely due to the difference in viscosity in the two materials and indicates that the two substances may be distinguishable via estimations of their dielectric properties. Interestingly, areas where there is low wind and areas where there is thin oil sheen are indistinguishable from the surrounding ocean after inversion. Both these phenomena show up in SAR images as dark patches similar to oil slicks but are of little concern to first responders of oil disasters. In summary, the PTSM has shown to provide reasonable estimates of the epsilon value of oil slicks in SAR imagery. Although portions of slick are slightly underestimated (unemulsified crude oil has an epsilon value of approximately 2.3) further avenues of research are currently ongoing in order to improve estimates.
14:20 -
On the Co-Cross Polarization Coherence over Sea Surface from Sentinel-1 TOPS Data
Longépé, Nicolas (1); Husson, Romain (1); Mouche, Alexis (2); Pottier, Eric (3); Archer, Olivier (2) - 1: CLS, France; 2: IFREMER, France; 3: University of Rennes I, France
SAR ocean surface wind retrieval has been originally based on a single observed quantity; the co-polarized Normalized Radar Cross Section (NRCS). This approach is directly derived from scatterometry and relies on a transfer function between radar observables and surface wind speed and direction, so called CMOD (C-band MODel) for C-band. In 2009, the launch of Radarsat-2 enabled to get routine measurements in cross-polarization. The weak sensitivity of the cross-polarized NRCS to incidence angle, to wind direction relative to the antenna look angle and the higher sensitivity to wind speed than for co-polarization are three major results that fostered many wind applications. The most striking is certainly the direct use of the cross-polarized channel for hurricane wind measurements. The capacities of Radarsat-2 also enable to measure the correlation between VH and VV channels. In particular, [Zhang et al. 2014] showed that the correlation between co- and cross- polarization channels has odd symmetry with respect to wind direction that could be complementary to the even symmetry of the co-pol NRCS. This work has been focused on the new capabilities of Sentinel-1 SARs, i.e. large swath acquisitions with phase preserving dual-pol channels. In this presentation/paper, the following topics will be detailed: - A massive processing of SLC IW products (several TB) collocated with wind model information has been carried out - First analysis of this database enables to outline some issues in the estimates of the co-cross coherence over ocean sea surface from S-1. Both the real part and the imaginary parts of the coherence should be equal to zero in the condition of reflection symmetry (up- or down-wind situation), but measurements show some small biases - We investigate then the reasons of these measured biases. Especially, the polarimetric distortion of S-1 sensor is neither estimated nor compensated on the produced data. - A methodology is proposed to account for this phenomena, enabling to counterbalance this effect and retrieve consistent results with those published in [Zhang et al. 2014]. - The analysis of the Pol-Calibrated co-cross coherence is carried out for a wide range of incidence angle and wind condition (which was not performed by [Zhang et al. 2014] due to limited availability of quad-pol RS-2). It shows interesting features which may be helpful to further understand the complex interactions between electromagnetic waves and sea surface. - The applicability to use this new information in the wind retrieval scheme is discussed. It is shown that S-1 co-cross coherence has burst-to-burst residuals which are not explained yet.
14:40 -
A MULTI-POLARIZATION ANALYSIS OF AZIMUTH CUT-OFF FOR SAR WIND SPEED RETRIEVAL UNDER MODERATE AND EXTREME WEATHER CONDITIONS
Corcione, Valeria (1); Nunziata, Ferdinando (1); Portabella, Marcos (2); Grieco, Giuseppe (3); Migliaccio, Maurizio (1) - 1: Università degli Studi di Napoli Parthenope, Italy; 2: The institute of Marine Sciences (ICM-CSIC), Spain; 3: Koninklijk Nederlands Meterologisch Instituut (KNMI), De Bilt, The Netherlands
Oceanic processes are driven by several key variables, e.g.; significant wave height (SWH), winds and sea state. Within this context, satellite microwave active remote sensing sensors are widely used to provide accurate sea surface information. The role of these sensors for ocean applications is worldwide recognized. In particular, the Synthetic Aperture Radar (SAR) can provide moderate-to-fine spatial resolution ocean surface products. At this aim, there exist Normalized Radar Cross Section (NRCS)-based approaches or spectral-based ones. In this work, the azimuth cut-off (λc) spectral method is investigated for both the co-polarized (VV) than cross-polarized (VH) channel. Originally, the azimuth cut-off technique was proposed in [1] to retrieve SWH, without the knowledge of any a priori information. The theory that is at the basis of the azimuth cut-off method is related to the influence of the orbital motion related to the surface waves in the SAR imaging of the ocean surface. This orbital motion results in additional Doppler shifts that distort the phase history of the backscattered signal that is used to synthetize the resolution in azimuth. The result is a low-pass filtered SAR image in the azimuth direction. The azimuth cut-off well correlates to the SWH because of its sensitivity to the long waves. Recently, in [2] the ACF-based λc approach has been improved to deal with high wind speed regimes, e.g.; extreme weather conditions. The key issues that allow to extend the method to high wind regimes concern the tuning of the method with respect to pixel spacing, box size and the homogeneity of the SAR imagery. In particular, the box size is set at about 1 km × 1 km and an adaptive window size is selected for the median filter to account for the pixel spacing and first results on wind speed estimation under tropical cyclone conditions are presented. In this study, Sentinel-1A dual- polarimetric (VV-VH) SAR data, collected in Interferometric Wide (IW) mode under moderate and extreme wind regimes (such as tropical cylcones), are used to investigate the sensitivity of azimuth cut-off to wind speed in both co- and cross-polarized channels. In particular, a comparison between the co-polarized and cross-polarized azimuth cut-off values is carried out, to further investigate the VH-cut-off map and to understand its information content. In addition, a coherent analysis based on the polarization entropy is carried out to understand the role of coherent information in estimating wind speed. Preliminary results, obtained processing a large data set of Sentinel-1 SAR measurements with co-located ECMWF ancillary wind speed information, show that the two azimuth cut-off maps (co- and cross- polarized) are well correlated when misfit and unreliable values are filtered out, while a slight negative correlation is found when comparing the entropy with the azimuth cut-off results. [1] [1] V. Kerbaol, B. Chapron, and P.W. Vachon, “Analysis of ERS-1/2 synthetic aperture radar wave mode imagettes”, Journal of Geophysical Reserarch, vol. 103, no. C4, pp. 7833-7846, 1998. [2] M. Portabella, V. Corcione, X. Yang, Z. Jelenak, P. Chang, G. Grieco, A. Mouche, F. Nunziata, W. Li, “Analysis of the SAR-derived wind signatures over extra-tropical storm conditions”, Dragon 4 Symposium, Copenhagen, Denmark, 26-30 June.
15:00 -
C-band Right-Circular Polarization Ocean Wind Retrieval
Zhang, Guosheng (1,2); Zhang, Biao (1); Perrie, Will (2); He, Yijun (1); Khurshid, Shahid (3); Warner, Kerri (3) - 1: Nanjing University of Information Science and Technology; 2: Bedford Institute of Oceanography, Canada; 3: Environment Climate Change Canada
In this study, we investigate the potential capability of ocean wind retrieval from C-band four polarizations (RV, RH, RR and RL) of the RCM CP mode and find that the RR-polarization is almost not sensitive to the wind direction, which is different from the other three polarizations. Based on the simulated RCM images and buoy measured winds, we propose a C-band RR-polarized wind retrieval model as a function of wind speed and radar incidence angle. The right hand circular polarizations in transmit and in receive are in opposite directions, as the transmit direction and receive direction are opposite. If we see the circular polarizations from the satellite, the circular is in clockwise direction when it transmits whereas it is in counter-clockwise direction when receives. If we decompose the RR-polarization to the vertical and horizontal directions, the RR-polarization can be a function of HV and VH polarizations. Therefore, C-band RR-polarization has the similarly linear relationship with the wind speed as the C-band linear cross-polarizations (VH/HV) in RADARSAT-2. For wind directions, the simulated NRCSs in the other three polarizations (RL, RV, RV and RH) have obvious dependences on the relative wind directions for a given wind speed and incidence angle (Figure 2). Based on interactions between radar microwave and sea surface waves generated by wind, Zhang et al. (2018) proposed an ocean semi-empirical model for RADARSAT Constellation Mission (RCM) RV and RH polarizations and suggested that RV-polarization is a better choice for ocean wind monitoring than RH-polarization. By analysis of the texture relationship between RCM CP mode and RADARSAT-2 linear quad-polarizations (VV, HH, VH and HV), Geldsetzer et al. (2015) demonstrated that the RV-polarized (or RH-polarized) NRCS approximately equals half of VV-polarized (or HH-polarized) NRCS, which is (σ_0RV≈0.5∙σ_0VV) or (σ_0RH≈0.5∙σ_0HH). Therefore, the VV-polarized (or HH-polarized) wind retrieval method can be employed for the RV-polarized (or RH-polarized) wind inversion. However, the fact is that no well-developed models exist for the HH-polarized wind retrieval, and the widely used approach has been to convert HH-polarized NRCS to VV-polarization using a polarization ratio and then retrieve winds with the VV-polarized function (Lin et al., 2008; Zhang et al., 2012). The RV-polarized wind retrieval model is supposed to be related to the VV-polarized function (Geldsetzer et al., 2015), which is CMOD5.N a well-developed VV-polarized wind retrieval function. In the presentation, we also show the RCM RV-polarized ocean semi-empirical model (Zhang et al., 2018) results for the same wind speed and incidence angle, as well as RV-polarized datasets, and that RV-polarization has a potential capability for the wind direction retrieval in the future. After all, the RCM CP quad-polarization will provide wider swath than RADARSAT-2 linear quad-polarized SAR images. The linear relationship between ocean wind speeds and C-band RR-polarized radar signal has important implications for oceanographic and meteorological researches, especially for tropical cyclone monitoring and dynamical studies. Based on the linear relationship between C-band VH/HV polarized NRCSs and wind speeds, we have studied the hurricane structures (Zhang et al., 2014, 2017), internal dynamic processes of eyewall replacement cycles (Zhang and Perrie, 2018, Remote sensing) and hurricane asymmetries (Zhang and Perrie, 2018, GRL). Associated with the advantages of three satellites in constellation, we would get two times more images in C-band RR-pol from RCM for a same tropical cyclone, which would improve our understanding of tropical cyclones.
15:20 - 15:50
15:50 - 16:10
15:50 -
Wind speed retrieval from simulated RADARSAT Constellation Mission (RCM) Compact Polarimetry (CP) SAR data for marine wind application
Khurshid, Shahid Khawaja (1); Perrie, William (2); Warner, Kerri (1); Geldzetser, Torsten (3); Zhang, Guosheng (2); Flett, Dean (1) - 1: Environment & Climate Change Canada, Canada; 2: Fisheries & Oceans Canada, Bedford Institute of Oceanography; 3: Independent Contractor
The operational National SAR winds (NSW) system derives marine wind speed estimates from RADARSAT-2 and Sentinel-1 data using published C-band models (CMOD) and provides these data and other products in near-real-time (NRT) to operational marine forecasters and meteorologists within Environment and Climate Change Canada (ECCC). The NSW system is being updated to process data from the future RADARSAT Constellation Mission (RCM), the primary objective of which is to ensure SAR data continuity and to improve operational capability of existing applications. A database is developed of simulated RCM Normalized Cross Sections (NRCS) generated from a Compact Polarimetry (CP) simulator using RADARSAT-2 quad-pol SAR images and co-located ocean wind vector observations from in-situ buoys. Twenty CP parameters and six standard polarization parameters are evaluated for their wind retrieval potential for three simulated RCM beam modes. The main purpose of this database is to develop and validate new wind retrieval models for RCM CP SAR data expected to be available to the operational NSW system in 2019. The simulated CP and standard parameters are analyzed for dependencies for wind speed, incidence angle and wind direction. Based on these analyses we recommend optimized RCM beam modes and polarizations and capabilities and limitations for ocean surface wind speed estimation. These analyses provide a refined reference database for wind retrieval models development and validation for NSW operational applications.
16:10 - 16:30
09:30 - 10:30
09:30 -
Exploring the Potential of TanDEM-X Dual-Polarization Time Series to Monitor Snow Accumulation on an Alpine Glacier
Schoenfeldt, Miriam; Parrella, Giuseppe; Hajnsek, Irena
Mass balance is a key indicator of the dynamics of a glacier. It is defined as the difference between the mass gained by accumulation and the mass lost due to ablation processes. Snow accumulation represents a major contribution to the accumulation term and its knowledge is therefore strictly required for the estimation of mass balance. So far, snow accumulation is being measured mainly by means of sparse weather stations which provide only point measurements. Snow maps for larger areas are obtained by spatially interpolating data from stations at different locations with the help of meteorological models. In the last decades, space-borne radar remote sensing has led to significant developments in the field of glaciology, due to the large spatial coverage with a relative short revisit time and a high spatial resolution. Thanks to the penetration capability of microwaves into dry snow and ice, conventional SAR systems are able to sense the surface as well as subsurface layers of glaciers and ice sheets, providing a tool to access information about the structure of the shallow snow-cover. This study investigates the potential of dual-polarimetric (HH/VV) TanDEM-X time series to monitor snow accumulation over the Aletsch glacier, in the Swiss Alps. In particular, co-polarization phase differences are employed to detect snowfall events, as shown already in [1] for the case of snow-covered soil. Furthermore, the propagation model described in [2] is adopted to link phase differences to fresh snow properties, like structural anisotropy, density and thickness, and to invert fresh snow depth from TanDEM-X data. Based on the inversion results, a simple algorithm has been developed to obtain an estimate of the temporal evolution of the snow height on the glacier surface. Preliminary results show that the proposed approach is able to provide a good estimate of the snow accumulation, as confirmed by snow depth measurements from a weather station nearby the glacier. However, a number of limitations remain, mainly related to the relatively low temporal sampling (11-day) offered by TanDEM-X and the need of some a priori knowledge about fresh snow density and structural anisotropy, which require further investigations. [1] Leinss S., Parrella G. and Hajnsek I.: Snow height determination by polarimetric phase differences in X-band SAR data, JSTARS, vol. 7, no. 9, pp. 3794-3810, 2014. [2] Leinss S., Loewe H., Proksch M., Lemmetyinen J., Wiesmann A. and Hajnsek I.: Anisotropy of seasonal snow measured by polarimetric phase difference in radar time series, The Cryosphere, vol. 10, pp. 1771-1797, 2016.
09:50 -
L- and P- band tomographic imaging in dense forests: AfriSAR results
EL Moussawi, Ibrahim (1,2,3,4); Ho Tong Minh, Dinh (1); Baghdadi, Nicolas (1); Abdallah, Chadi (2); Jomaah, Jalal (3); Strauss, Olivier (4); Lavalle, Marco (5) - 1: Irstea UMR TETIS University of Montpellier (UM); 2: CNRS-L; 3: Lebanese University (UL); 4: Lirmm University of Montpellier (UM); 5: NASA Jet Propulsion Laboratory JPL California Institute of Technology USA
This study presents tomographic analysis using L-band NASA/JPL UAVSAR and P-band ONERA/SETHI from AfriSAR data conducted over the Gabon Lope Park. The objective of this paper is to provide a better understanding of tomographic capabilities in characterization of dense forested areas at L-band. Prior to tomographic imaging, a phase residual correction methodology based on Sum Kronecker Product and Phase Double Localization PCDL have been implemented. The estimated vertical structure of the forest extracted from the correct tomographic data is validated with small footprint light detection collected during the AfriSAR campaign in July 2015. The results show that L-band, similar to P-band, allows to retrieve the whole forest structure and characterizes the ground and volume scattering. We demonstrate that L-band tomographic imaging can now be carried out even in dense tropical forest.
10:10 -
POLARIZATION EFFECT ON GLACIER MOVEMENT ESTIMATION USING DIFFERENTIAL SAR INTERFEROMETRY (DInSAR)
Khati, Umesh
Nela, Bala Raju; Singh, Gulab; Khati, Umesh - Centre of Studies in Resources Engineering, IIT Bombay, India
Glaciers are the most important component in the cryosphere and it’s a reliable indicator of climate change. We can understand these climate changes by monitoring the glacier dynamics. Glacier velocity is one of the important parameters to know about glacier dynamics and it’s health. This glacier surface velocity is further useful to calculate the glacier thickness and mass balance. Remote sensing techniques especially radar remote sensing techniques are cooperating to study the glacier dynamics. Differential SAR Interferometry (DInSAR) is the radar interferometry technique to measure surface movement with an accuracy of millimeter range by differencing two Interferograms. The first time this DInSAR technique was used for cryosphere application in 1994 by Goldstein to monitor the Rutford Ice Stream motion, Antarctica. All the previous DInSAR glacier movement studies have been done with the HH polarization only. But the mountain glaciers are irregular in structure and shape and moreover, it covered with different kinds of materials like dry/wet snow, sediments, moraines and boulders. Therefore it's not always correct to use only HH channel SAR images to estimate glacier movement using the DInSAR technique. This DInSAR technique will give better and extra information with the help of polarimetry because it contains the backscattered information of target point geometrical structure, shape and orientation. The first time we used this DInSAR technique with 4 polarizations for ‘glacier movement’ to select an optimum pair. We selected Bara Shigri glacier, the largest glacier in the Indian state of Himachal Pradesh of a length nearly 28 km and area 126.5 km2 and it contains glacier ice, debris as well as a medial & lateral moraine. About 14% of this Bara Shigri glacier surface is extensively covered with debris. The less temporal baseline and high wavelength of SAR images give better results for glacier movement and we clearly observed this in coherence and interferogram images. This coherence value varies in between 0 to 1 and it indirectly tells the quality of the DInSAR. If this coherence value is more, then the results will be more accurate and if it's less than 0.3, will not be useful for the DInSAR process. But it’s impossible to get the high coherence value for the continuous moving targets like a glacier. We compared the coherence images generated with the 4 different types of polarizations. Co-polarized ( HH and VV) giving the same and high coherence for the snow-covered area. Cross-polarized ( HV and VH) giving the same and high coherence for the debris-covered area. Therefore, selecting of polarization channel for DInSAR process is dependent on the type of glacier. If the glacier is completely covered with snow, better to use co-polarized and for the debris-covered glacier, it would be better to use cross-polarized. The maximum velocity of a Bara Shigri glacier we observed 15.3 cm/day in the accumulation region, near to the portion of the medial moraine in the month of accumulation season (March 2015). We can also observe time series changes in a glacier by comparing recent velocity maps (using Sentinel-1A/1B, ALOS-2 data) with old velocity maps (using ERS tandem pair images). For comparison purpose, better to use the same bands if it’s giving good coherence. Using Laminar flow equation we can also derive thickness from the velocity and then we can calculate glacier mass balance. If this two thickness and mass balance are giving accurate results, then this DInSAR technique will be the best method to know about glacier dynamics using the glacier movement.
10:30 - 11:00
11:00 - 12:00
11:00 -
Polarimetric Sensitivity of Multi-Frequency Airborne SAR Measurements to the Ice Zones of Greenland
Parrella, Giuseppe (1); Papathanassiou, Kostas (1); Hajnsek, Irena (1,2) - 1: German Aerospace Center (DLR), Germany; 2: ETh Zurich, Switzerland
Polarimetric SAR data are well known to provide a better characterization of a scattering scene compared to single-polarization measurements. In order to further extend the observation space, multi-frequency (polarimetric) measurements can be considered. The joint use of different wavelengths allows to gain sensitivity about scatterers at different size scale and at different depth (in case of volumes) due to the different penetration depths. For the study of ice masses, multi-frequency Pol-SAR data are expected to be sensitive to different subsurface layers. For instance, X-band is more suitable to investigate the shallow snow cover as well as surface features while L-band allows to sense the underlying firn and ice layers. Therefore, such a dataset can potentially provide a significant contribution to the identification and characterization of different ice zones which, in turn, is needed for more accurate mass balance estimation. Because of the large number of factors determining the subsurface structure of glaciers and ice sheets, the exploitation of PolSAR data still plays a secondary role in the study of snow and ice properties. Early studies addressed the identification of ice zones focusing on the analysis of backscattering coefficients in single [1] and multi-polarization configurations [2]. Recent studies have shown that also the coherent nature of the polarimetric signature is essential. For instance, it has been shown that polarimetric phase differences between the HH and VV channels can reveal details of the microstructure of snow and firn layers [3]. In this study, a multi-frequency analysis of polarimetric signatures over the different ice zones of the Greenland ice sheet is presented. A set of descriptors is employed to extract and interpret the polarimetric information from the data, which includes backscattering coefficients, the scattering entropy, the mean alpha angle, polarimetric ratios and phase differences. The study is based on a multi-frequency (L-, C- and X- band) airborne Pol-SAR dataset acquired in May 2015, during the ARCTIC15 campaign, over a 200 km long (and 5 km wide) transect in West Greenland. Preliminary results show that, X- and C-band signatures are dominated by the ice surface features (e.g. roughness), which can generate complex (volume-like) scattering at such short wavelengths even in the bare ice (ablation) zone. Over the firn (accumulation) zone, the polarimetric signatures are saturated due to strong volume scattering occurring within the firn layers. In contrast, L-band measurements show a pronounced change of polarimetric signatures over the area of transition between two zones, pointing out a clear variation of scattering mechanisms according to the different subsurface structure. [1] M. A. Fahnestock, R. Bindschadler, R. Kwok, and K. C. Jezek, “Greenland ice sheet surface properties and ice dynamics from ERS-1 SAR imagery,” Science, 262(5139), 1530-1534, Dec. 1993. [2] M. Koenig, J.-G. Winther, N. T. Knudsen and T. Guneriussen, Equilibrium- and firn-line detection with multi-polarization SAR – first results, Proc. EARSel-SIG Workshop Land Ice and Snow, Dresden, Germany, 16-17 June, 2000. [3] Parrella G., Hajnsek I. and Papathanassiou K.: On the interpretation of polarimetric phase differences in SAR data over land ice, GRSL, vol. 13, no. 2, pp. 192-196, 2016.
11:20 -
Ice thickness change in Indian Himalayas using SRTM and TanDEM-X global DEM
Khati, Unmesh
Bandyopadhyay, Debmita; Singh, Gulab; Khati, Unmesh - Indian Institute of Technology Bombay, India
The Himalayas form one of the largest glacier systems in the world and have immense stored water potential. In fact, it is called the ‘third pole’ as well as the ‘water tower of Asia’. However, owing to the changing climatic conditions, the rate of retreat of majority of the glaciers in India have accelerated. Hence, there is a need for constant monitoring of these freshwater reservoirs. With the dearth of ground-based information in these rugged terrains, remote sensing technology is the most sensible way to gauge the changes taking place in the entire range of the Himalayas. Radar remote sensing has an edge over other techniques (like optical) in terms of all-weather and all-day capability. Using interferometric SAR (InSAR), Digital Elevation Models (DEMs) have been generated which are able to give an idea about the topography of the region in great detail. Moreover, with the recent release of TanDEM-X DEMs (90m) in 2018, observing decadal changes has become readily achievable. Further, with TanDEM-X being the only global DEM generated using X-band data after SRTM in the year 2000, which is disseminated freely, such dataset at our disposal has tremendously improved the scope of interferometry based studies in the glaciated terrain. Glaciers can be monitored using a variety of parameters like change in length, area, thickness in order to understand the change in glacier dynamics in response to the changing climate. This can be corroborated with parameters like temperature or precipitation patterns. Of all the glacial parameters, mass balance has been considered as one of the most responsive and representative of estimates. For mass balance, there are various methods that have been deployed like energy balance, geodetic and ELA/AAR method. Geodetic method utilizes the information of thickness and density change, which is heavily dependent on the accuracy of DEMs. With the recent release of TanDEM-X global DEM, in this study, we not only try to estimate the thickness change in the entire Indian Himalayan terrain but also try to highlight the potential of the TanDEM-X DEM, to monitor such dynamic geographical features. Hence, this study would facilitate assessing the water sustainability potential and the contribution of the Indian Himalayas towards the melt-water for the early twenty-first century, which will eventually help quantify the sea-level rise at a global scale.
12:00 - 12:20
12:20 - 13:20
13:20 - 13:40
13:20 -
Investigating the Potential to Retrieve Vertical Subsurface Structures of Ice Sheets by Means of Pol-InSAR Data
Fischer, Georg (1,2); Papathanassiou, Konstantinos (1); Hajnsek, Irena (1,2) - 1: German Aerospace Center (DLR), Germany; 2: ETH Zurich, Institute of Environmental Engineering, Switzerland
Polarimetric SAR interferometry (Pol-InSAR) techniques allow investigating the vertical distribution of scattering processes, and have therefore the potential to provide geophysical information about the subsurface structure of glaciers and ice sheets. Studies have shown the retrieval of InSAR penetration depths [1] and extinction coefficients [2] at different frequencies and polarizations, which are linked to the geophysical subsurface characteristics. The associated retrieval schemes, e.g. by means of a Random Volume under Ground model [2], are based on the assumption of a lossy propagation through an isotropic, homogeneous volume in the subsurface of glaciers or ice sheets. This is described by a constant signal extinction coefficient, which results in an exponential decrease of backscatter intensity along depth. However, recent studies [3] have shown that the constant extinction assumption is insufficient to describe airborne Pol-InSAR data from Greenland, due to the presence of heterogeneously distributed scatterers within the firn volume. Refreezing of melt water leads to distinct horizontal ice layers in the subsurface, which have a clear effect on the backscattered signal. These layers can be modeled as Dirac deltas with a given backscattering power at a given depth. In addition, backscattering originates also from the firn body and ice inclusions distributed across depth, which can be described with vertical volume structure functions. While the assumption of a constant extinction for the volume is accurate enough to describe InSAR coherence magnitudes, it fails when simulating the interferometric phase in our data [3]. Non-constant extinction models are therefore necessary and preliminary results show that only one additional degree of freedom in the volume structure function significantly improves the agreement with the data. The higher modeling complexity, introduced by additional parameters for distinct subsurface layers and a volume structure with non-constant extinction, increases the challenge of a retrieval or inversion scheme. The inversion depends strongly on the observation scenario as well as on the subsurface characteristics of the area under investigation. A simple case, which roughly represents the subsurface of our test site in the percolation zone of Greenland, consists of two layers and a volume, where the volume is independent of the polarization (i.e. random volume). The resulting eleven model parameters could be theoretically inverted with a dual-baseline Pol-InSAR setup providing twelve independent observables. Any additional layer or a polarization dependent volume (i.e. oriented volume) add at least two parameters and will complicate the inversion. This study investigates the limitations and potential of polarimetric and (multi-baseline) interferometric data for the retrieval of ice sheet subsurface structures and performs a sensitivity study based on different modeling complexities. Polarimetry helps to separate the effects of distinct layers, assuming rough surface scattering, and the volume. With a limited amount of interferometric baselines, potential lies in the combination of data with small vertical wavenumber, which are more affected by the volume, and data with larger vertical wavenumber, which are more affected by the layers. [1] E. W. Hoen and H. Zebker, “Penetration depths inferred from interferometric volume decorrelation observed over the Greenland ice sheet,” IEEE Trans. Geosci. Remote Sens., vol. 38, no. 6, pp. 2572–2583, Nov. 2000. [2] J.J. Sharma, I. Hajnsek, and K.P. Papathanassiou, “Estimation of glacier ice extinction using long-wavelength airborne Pol-InSAR,” IEEE Trans. Geosci. Remote Sens., vol. 51, no. 6, pp. 3715-3732, Jun. 2013. [3] G. Fischer, G. Parrella, K. P. Papathanassiou, and I. Hajnsek, “Sensitivity of polarimetric SAR interferometry data to different vertical subsurface structures of the Greenland ice sheet,” in Proc. of IGARSS, Forth Worth, USA, 2017, pp. 3581–3584.
13:40 - 14:00
14:00 - 14:40
14:00 -
Analysis of PolSAR imagery from the DALOEX 2015 technology demonstration campaign in Greenland
Krogager, Ernst - Danish Defence Acquisition and Logistics Organisation, Denmark
As a technology demonstration related to future capabilities in the Arctic region, the Danish Defence Acquisition and Logistics Organization (DALO) conducted a test campaign, DALOEX 2015, in Greenland from late April to late May 2015, where an airborne multiband, high-resolution, fully polarimetric synthetic aperture radar (PolSAR) system was flown over several test areas. The F-SAR system of the German Aerospace Center (DLR) was chosen for the test campaign and provided the desired possibility of exploring five frequency bands in the range of 400 MHz to 10 GHz with fully polarimetric data acquisition. Polarimetric radar imaging requires more complex and more costly systems, but also more complex methods for visualization, interpretation and processing. For such reasons, fully polarimetric radar systems are still not commonly used for military applications, but nevertheless, the potential benefits of fully polarimetric systems must be taken into due consideration for future capabilities. Hence, a main objective of the test campaigns reported here was to demonstrate and illustrate how the utilization of information carried by the polarization of electromagnetic waves could improve the performance of imaging radar systems. The test scenarios included experiments with objects hidden under snow as well as moving targets and boats near icebergs. Detailed ground truth was collected in the form of precision GPS measurements with associated photos, and aerial photos were taken during helicopter flights. In this presentation, examples of results and findings are presented with a focus on methods for interpretation and visualization of multiband PolSAR imagery. Many such methods have been devised and developed since the early days of radar polarimetry in the 1950s. Notably, polarimetric decompositions have been developed for characterizing targets in terms of physical scattering mechanisms and mathematical formulations on a pixel by pixel basis. Polarimetric decompositions are usually categorized in two main groups: coherent and incoherent methods, the latter implying the use of averaged quantities and statistical concepts. Incoherent approaches seem to be preferred in many cases, but it should be kept in mind that SAR is coherent in nature, relying on a coherent integration of signal samples along the path of the radar platform. Coherent decompositions aim at extracting target characteristic features for each individual (complex-valued matrix) pixel of the raw SAR image before performing any averaging, while incoherent approaches are typically based on forming coherency or covariance matrices before the decomposition. A variety of interpretation schemes have been proposed and widely applied in modern SAR polarimetry. However, such interpretations are to a large extent based on assumptions and approximations in order to extract features associated with particular scattering phenomena and target structures. The target scattering matrix for each pixel effectively contains five independent parameters: three amplitudes and two relative phase terms. The three amplitudes can be used for generating color images by letting the three components modulate the RGB (red, green, blue) color pixel by pixel, but the use of polarimetric decompositions makes it possible to form alternative RGB images with colors more closely related to basic scattering mechanisms. A main challenge is to extract and separate as many target characteristic features as possible under the limitation of five independent parameters for each pixel. Generic scattering objects are commonly used for such characterizations, e.g., spheres, dihedrals, dipoles, wires, helices, quarter-wave plates, and in the case of incoherent decompositions, concepts like volume scattering and rough surface scattering are used to characterize targets and extended target areas on a statistical basis. The classical Pauli decomposition known from mathematical physics is a fundamental decomposition used in the outset for many decompositions in radar polarimetry. In relation to radar targets, the Pauli decomposition represents an odd-bounce reflector and two even-bounce reflectors oriented at 0 and 45 degrees, respectively. One disadvantage of this representation is due to the fact that an arbitrarily oriented dihedral produces contributions to two out of three components, which renders the interpretation of a three-component (RGB) color image difficult. An alternative approach, the socalled sphere, diplane, helix decomposition, was proposed by the author in 1990. According to this approach, a scattering matrix is decomposed into an odd-bounce reflector, an even-bounce reflector and a helix. A major advantage of this approach is that three fundamentally different object types contribute to one and only one component, unlike other approaches, which suffer the same ambiguity problem as the Pauli matrix representation. A disadvantage, which has been pointed out in other works, is that the diplane and helix components are not mutually orthogonal and hence not satisfying the criteria of many statistical approaches and mathematical transformations. In spite of the orthogonality issue, the quantities of the sphere, diplane, helix decomposition are closely related to how a radar sees the reflection from a complex scatterer. Thus, the components can be measured directly in the circular polarization basis, independent of the incidental overall orientation angle of the target around the line of sight. From this point of view, the reflection from any complex structure appears to the radar as if it were due to the combined response from one sphere, one diplane and one helix. A dipole appears as half sphere and half diplane, and it is not possible without additional information or a priori knowledge to determine, whether the target is actually a dipole. This is just one example of the fundamental limitations of polarimetric radar target characterization and at the same time an illustration of how attempts to extract more information than is actually contained in the data may lead to confusion and misrepresentations. With due regard to the objective of assessing the utility of fully polarimetric radar systems for applications in the Arctic (and elsewhere), the above aspects will be further addressed in the presentation and illustrated by examples from DALOEX 2015 as well as from a preparatory campaign (DALOEX 2014) with F-SAR in Denmark in October 2014.
14:20 -
DLR Airborne SAR Campaign on Permafrost Soils and Boreal Forests in the Canadian Northwest Territories, Yukon and Saskatchewan: PermASAR
Hajnsek, Irena (1,2); Joerg, Hannah (1); Horn, Ralf (1); Keller, Martin (1); Gesswein, Daniel (1); Jaeger, Marc (1); Scheiber, Rolf (1); Bernhard, Philipp (2); Zwieback, Simon (2) - 1: Microwaves and Radar Institute DLR, Germany; 2: Institute of Environmental Enginnering ETH Zürich
Monitoring permafrost regions from space is one of the objectives of the Tandem-L mission foreseen to be launched in the near future. In times of climate change and global warming the permafrost boundary moves north more rapidly every year. Permafrost soils contain CO2 and even worse methane (CH4). When thawed (or activated) these critical gases are released to the atmosphere [1]. Ahead of the Tandem-L satellite mission the DLR Microwaves and Radar Institute (DLR-HR) conducted an experimental airborne SAR campaign involving its multi-frequency and polarimetric F-SAR instrument in the Canadian Northwest (Northwest Territories and Yukon) and Saskatchewan. The first campaign took place in July/August 2018 to cover the thawed regions and a second campaign will take place in March/April 2019 covering the frozen state of the soil surfaces. Ten test sites have been selected representing a south-north gradient and covering different permafrost types. The main motivation for this campaign is to gain knowledge about the interaction between longer EM wavelength and permafrost soils (vegetated and non-vegetated). This is important in order to develop algorithms for the observation of changes in such regions, which in turn is important for the quantification of changes due to global warming. For the airborne campaign specific science questions have been drafted depending on soil type and vegetation coverage. The focus of this study will be to develop and validate algorithms on soil moisture, characterize organic layers, regions with different active layer thicknesses and characterize the vegetation on top of the soil [2,3]. In this paper the test sites along the south-north gradient are presented and their responses at different frequencies assessed. In addition their polarimetric and interferometric behaviors are analyzed. In summary a first assessment of the data collected shall be presented. [1] Zwieback S, Kokelj SV, Gunther F, Boike J, Grosse G, Hajnsek I. Sub-seasonal thaw slump mass wasting is not consistently energy limited at the landscape scale. Cryosphere. 2018;12(2):549-64. [2] Pichierri M, Hajnsek I, Zwieback S, Rabus B. On the potential of Polarimetric SAR Interferometry to characterize the biomass, moisture and structure of agricultural crops at L-, C- and X-Bands. Remote Sensing of Environment. 2018; 204:596-616. [3] Brancato V, Hajnsek I. Analyzing the Influence of Wet Biomass Changes in Polarimetric Differential SAR Interferometry at L-Band. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing. 2018;11(5):1494-508.
14:40 - 15:00
15:30 - 15:40
